Is This Video Real?

The Principal
Mischief of Deepfakes and How the
Lanham Act Can Address It
QUENTIN J. ULLRICH *

This Note argues that the false association cause of action under Section
43(a)(1)(A) of the Lanham Act is well-suited for addressing problems posed
by deepfakes, and outlines for practitioners the mechanics of such a cause
of action. A “deepfake,” which is a portmanteau of “deep learning” and
“fake,” is a digitally manipulated, often highly realistic video that
substitutes the likeness of one person with that of another. Due to the way
they deceive their viewers, deepfakes pose a threat to privacy, democracy,
and individual reputations. Existing scholarship has focused on
defamation, privacy tort, copyright, regulatory, and criminal approaches to
the problems raised by deepfakes. These legal approaches may at times be
successful at penalizing the creators of pernicious deepfakes, but they are
not based on a theory of consumer confusion, which this Note argues is the
principal mischief posed by deepfakes. Further, since deepfakes are often
uploaded anonymously and the only effective remedy is against website
owners, certain of these approaches are frustrated by the Communications
Decency Act’s immunization of website owners from liability for torts with
a “publication” element. Hence, this Note proposes that the law of false
association, which is principally concerned with consumer confusion, is best
suited for addressing deepfakes. Importantly, a Lanham Act cause of action
would allow victims of deepfakes to sue website owners under a theory of
contributory infringement, because the Communications Decency Act does
not immunize website owners from intellectual property claims.

* Notes Editor, Colum. J.L. & Soc. Probs., 2021–2022. J.D. Candidate 2022,
Columbia Law School. The author is grateful to Professor Jane C. Ginsburg and Professor
Laura A. Heymann for their thoughtful advice, the assiduous JLSP team for their edits and
comments, and an old friend for first teaching him about deepfakes. All errors are those of
the author.
2 Columbia Journal of Law & Social Problems [55:1

INTRODUCTION

It is October of an election year, and there is an inflammatory

video circulating of one of the presidential candidates criticizing
voters in key swing states. 1 Is it real, or is it a deepfake?
Deepfakes are digitally manipulated videos that substitute the
likeness of one individual with that of another, or otherwise alter
someone’s likeness. 2 These videos, often uploaded to the internet
anonymously, can appear highly realistic due to advances in video
editing techniques, machine learning, and artificial intelligence.
As a result, deepfakes have a high propensity to mislead their
viewers. Although prominent social media sites such as Twitter
have implemented policies forbidding misleading deepfakes, 3
other, more fringe, websites may either willfully ignore pernicious
deepfakes or encourage their circulation. Even if a video is
identified as fake, some websites may refuse to take the video
down, and the target of the video—such as the hypothetical
presidential candidate discussed above—may choose to resort to
the courts for compensation, an injunction, or greater public
recognition of the fakery.
This Note argues that a false association cause of action under
Section 43(a)(1)(A) of the Lanham Act is well-suited for addressing
deepfakes due to its focus on confusion and the fact the
Communications Decency Act (CDA) does not block false
association claims against website owners. 4 Section 43(a)(1)(A)
creates a cause of action for one whose likeness is used by another
in a way that is “likely to cause confusion, or to cause mistake, or
to deceive as to the affiliation, connection, or association of such

1. Deepfakes did in fact appear during the 2020 U.S. election. Fortunately, they were
generally either perceptibly fake, identified as fake, or otherwise had a relatively
undetectable impact on the outcome of the election. See Tom Simonite, What Happened to
the Deepfake Threat to the Election?, WIRED (Nov. 16, 2020), https://www.wired.com/story/
what-happened-deepfake-threat-election/ [https://perma.cc/LVG9-RUPR].
2. This Note uses the term “deepfake” broadly to refer to digitally manipulated videos,
whether they are low-quality or high-quality.
3. Synthetic and Manipulated Media Policy, TWITTER, https://help.twitter.com/en/
rules-and-policies/manipulated-media [https://perma.cc/ZKS9-MRC5].
4. See 47 U.S.C. § 230(e)(2). This Note uses the terms “false association” and “false
endorsement” interchangeably. Although courts often refer to the claim as “false
endorsement,” the term “endorsement” does not appear in Section 43(a) of the Lanham Act;
Section 43(a) does, however, use the term “association.” See 15 U.S.C § 1125(a)(1)(A).
Section 1125(a) of the Lanham Act is popularly known as Section 43(a). 1 CHARLES E.
MCKENNEY & GEORGE F. LONG III, FEDERAL UNFAIR COMPETITION: LANHAM ACT 43A § 1:1
(2021).
2021] Is This Video Real? 3

person with another person.” 5 This Note argues the hypothetical

presidential candidate described above, or a similarly harmed
individual, can sue the creator of a confusing deepfake using a
theory of false association under Section 43(a)(1)(A). Additionally,
and crucially, because the individual who posted the deepfake may
be unidentifiable and therefore incapable of providing relief, this
Note argues a plaintiff could pursue a claim of contributory false
association against the owner of a website hosting confusing
deepfakes. Unlike defamation claims, which are barred by Section
230 of the Communications Decency Act, injured parties may bring
false association claims not only against direct actors, but also
against intermediary website operators who knowingly host the
misleading content. 6
This Note begins in Part I by explaining deepfakes and
categorizing them based on how the Lanham Act can address
them. Next, in Part II, this Note discusses the law of false
association under Section 43(a) of the Lanham Act and why it is
better suited than alternative legal strategies for addressing
deepfakes. Finally, in Part III, this Note discusses the mechanics
of a false association claim involving a deepfake, focusing on the
issues most likely to present a hurdle to such a claim.

I. DEEPFAKES AND THEIR DANGERS

While certain deepfakes pose harm to society, not all do.

Accordingly, the law should distinguish the good from the bad,
balancing enforcement with protecting free expression. After
explaining how deepfakes are created and distributed in Part I.A.,
this Note will provide a taxonomy of deepfakes in Part I.B,
distinguishing between (i) confusing and, therefore, harmful
deepfakes the Lanham Act should address, (ii) non-confusing but
harmful deepfakes the Lanham Act cannot address, and (iii) non-
harmful deepfakes the Lanham Act should not address.
Afterwards, in Part I.C., this Note will discuss potential non-legal
solutions to confusing and pernicious deepfakes, before turning to
legal solutions in Part II.

5. See 15 U.S.C § 1125(a)(1)(A).

6. See 47 U.S.C. § 230(e)(2); Gucci Am., Inc. v. Hall & Assocs., 135 F. Supp. 2d 409,
413 (S.D.N.Y. 2001) (holding that the Communications Decency Act does not immunize
website owners from trademark infringement claims, because they are intellectual property
claims explicitly excluded from the act under Section 230(e)(2)).
4 Columbia Journal of Law & Social Problems [55:1

A. DEEPFAKES 101

Deepfakes are often created using Generative Adversarial

Networks (GANs), which combine generative and discriminative
algorithms. 7 Generative algorithms create data based on certain
inputs, while discriminative algorithms classify data based on
certain inputs. 8 In the case of GANs, both generative and
discriminative algorithms take as their inputs large datasets of
images, sounds, or videos and work together to create a new image,
sound, or video, which approximates those in the dataset but which
is not a direct copy of them. 9 The generative algorithm creates new
samples, which are then tested by the discriminative algorithm to
see if they appear “real” given the input data. 10 As GANs become
more sophisticated or use larger data sets as inputs, deepfakes will
appear more and more realistic. 11 Since the quality of a deepfake
depends on the amount of source material available, individuals
whose images and voices are widely available to the public are
more susceptible to being imitated via deepfakes. 12
Due to the ease of sharing content on social media, deepfakes
and the information (or misinformation) they convey can spread
swiftly. Consider the example of a distorted video, posted in 2019,
of Speaker of the House, Nancy Pelosi, which made her speech
sound slurred. The video, originally posted on the Facebook page
“Politics WatchDog,” was shared 45,000 times and viewed over two
million times within just one day of its original posting and
Rudolph Giuliani’s inclusion of it in a Tweet. 13

7. Mika Westerlund, The Emergence of Deepfake Technology: A Review, 9 TECH.

INNOVATION MGMT. REV. 39, 40–41 (2019).
8. Russell Spivak, “Deepfakes”: The Newest Way to Commit One of the Oldest Crimes,
3 GEO. L. TECH. REV. 339, 342–43 (2019).
9. See Westerlund, supra note 7, at 41.
10. Id. See also Chris Nicholson, PATHMIND, A Beginner’s Guide to Generative
Adversarial Networks (GANs), A.I. WIKI, https://wiki.pathmind.com/generative-
adversarial-network-gan [https://perma.cc/29E5-SRCK] (“One neural network, called the
generator, generates new data instances, while the other, the discriminator, evaluates them
for authenticity[.]”).
11. See Bobby Chesney & Danielle Citron, Deep Fakes: A Looming Challenge for
Privacy, Democracy, and National Security, 107 CALIF. L. REV. 1753, 1760 (2019).
12. See Westerlund, supra note 7, at 43.
13. Drew Harwell, Faked Pelosi Videos, Slowed to Make Her Appear Drunk, Spread
Across Social Media, WASH. POST. (May 24, 2019), https://www.washingtonpost.com
/technology/2019/05/23/faked-pelosi-videos-slowed-make-her-appear-drunk-spread-across-
social-media/ [https://perma.cc/UG5D-RRKQ].
2021] Is This Video Real? 5

Those creating deepfakes may be individuals or foreign

governments with political agendas, 14 businesses promoting their
products by using others’ likenesses, the hosts of YouTube
channels or Facebook pages seeking to generate viewership, or
trolls—online provocateurs without a particular agenda—taking
satisfaction in deceiving the public. 15 Importantly, those initially
posting deepfakes to social media are often careful to ensure their
anonymity by veiling their IP addresses. 16 As will be discussed in
Part II, the anonymity and extra-territoriality of deepfake
uploaders can present hurdles to lawsuits against them.

B. A TAXONOMY OF DEEPFAKES

This Note contemplates three categories of deepfakes: (i)

deepfakes that are harmful due to their ability to confuse viewers,
(ii) deepfakes that are not confusing but are harmful to individuals’
dignity and privacy, and (iii) deepfakes that are non-harmful or
even socially productive. These three categories, which may at
times overlap, are expounded in turn below. 17

1. Confusing Deepfakes

As the Speaker Pelosi example indicates, even low-quality

deepfakes can confuse viewers and distort their perceptions of
reality. In one of the first known instances of a political party
employing deepfakes, Belgium’s “sp.a” party released a deepfake
of President Donald Trump mocking the Belgian government for

14. Researchers from Purdue and Princeton University, for example, have cataloged
the growing threat of misinformation campaigns by foreign governments, noting seventy-
six influence campaigns between 2013 and 2019. DIEGO A. MARTIN ET AL., EMPIRICAL STUD.
OF CONFLICT PROJECT, TRENDS IN ONLINE INFLUENCE EFFORTS 1 (2020),
https://esoc.princeton.edu/publications/trends-online-influence-efforts [https://perma.cc/
KMD9-LNVU].
15. See Westerlund, supra note 7, at 41–42.
16. See Chesney & Citron, supra note 11, at 1792.
17. These categories set the basic parameters for this Note’s assessment of legal
remedies with respect to deepfakes, which are discussed in Part II and Part III. As will be
discussed in Part II, the Lanham Act can be used to hold accountable the creators of and
website hosts of confusing deepfakes, i.e., those falling in category (i). Correspondingly,
Lanham Act caselaw addressing First Amendment defenses can be used to ensure the
creators of socially productive—i.e., category (iii)—deepfakes are protected from false
association claims, as discussed in Part III. However, the Lanham Act cannot be applied to
deepfakes that do not confuse their viewers. As will be explained in Part II, the intentional
infliction of emotional distress tort and the right of publicity are likely best suited for
category (ii) deepfakes that harm not by confusion but by insult to dignity.
6 Columbia Journal of Law & Social Problems [55:1

staying in the Paris climate agreement. 18 The purported intent of

the deepfake was to motivate domestic political action on climate
change, and sp.a stated, “It is clear from the lip movements that
this is not a genuine speech by Trump.” 19 Many viewers in
Belgium, however, thought the video was real and expressed
outrage at the idea of an American president interfering with
Belgian policy. 20
Deepfakes have also emerged in the United States as a means
for pursuing political goals. Professors Bobby Chesney and
Danielle Citron, for example, have documented a viral deepfake
used to undermine the gun control movement. 21 In the aftermath
of the February 2018 school shooting in Parkland, Florida,
students who survived became national voices in the gun-control
movement. 22 One such national voice is Emma González, who was
featured in a gun-control advocacy video, in which she rips a piece
of paper featuring a bulls-eye target. 23 Soon after the video was
released, someone created a deepfake showing González ripping
not a bulls-eye but a copy of the United States Constitution. 24
Fortunately, the harm of the video was mitigated by the
availability of an original clip of González tearing the bulls-eye, the
fact that the video was poorly edited, and the intense media
scrutiny of events following Parkland. 25 Nonetheless, it is easy to
imagine a situation where an original video is unavailable, the
deepfake is higher quality, or media scrutiny comes too late to
rectify the harm.
Incidents like these forebode even more perilous uses of
deepfakes. On the evening before an election, rival political
parties, foreign state actors, or internet trolls could post deepfakes
of presidential candidates engaging in lewd acts or criticizing

18. See Oscar Schwartz, You Thought Fake News Was Bad? Deep Fakes Are Where
Truth Goes to Die, GUARDIAN (Nov. 12, 2018), https://www.theguardian.com/technology/
2018/nov/12/deep-fakes-fake-news-truth [https://perma.cc/Z8SB-FF2W]; Westerlund, supra
note 7, at 43. The party’s name, “sp.a,” was later changed to “Vooruit.” Socialist party SP.A
to change name to Vooruit, THE BULLETIN (Sep. 15, 2020), https://www.thebulletin.be/
socialist-party-spa-change-name-vooruit [https://perma.cc/PF27-PC7F].
19. Schwartz, supra note 18.
20. Id.
21. See Chesney & Citron, supra 11, at 1755–57.
22. Id. at 1755.
23. Id. at 1755–56.
24. Id. at 1756.
25. Id.
2021] Is This Video Real? 7

voters in a swing state. 26 Indeed, Untied States government

agencies and intelligence organizations have expressed concern
over how deepfakes could be used to undermine national
security. 27
Numerous other risks posed by deepfakes have been
extensively documented by other scholars. For example, criminals
can create deepfakes for the purposes of blackmail, or terrorist
organizations can disseminate deepfakes depicting violence
against them in order to garner support. 28 Rogue investors can
unleash deepfakes in order to manipulate markets. 29
The common ground of all of these examples is confusion—due
to how realistic the video looks, the unapprised viewer believes the
deepfaked individual is engaging in or endorsing the fake actions
or statements that are depicted.

2. Non-Confusing but Harmful Deepfakes

Although this Note focuses on deepfakes that cause harm by

way of confusing their viewers, deepfakes can be harmful even
without generating confusion. A common example is pornographic
deepfakes. Pornographers (and even amateur video editors) have
created deepfakes depicting celebrities engaging in sex acts, and
software programs allow users to automatically swap others’ faces
into sex videos. 30 Social media sites like Reddit and porn sites like
Pornhub have banned deepfake sex videos; 31 other sites like

26. Nina I. Brown, Deepfakes and the Weaponization of Disinformation, 23 VA. J.L. &
TECH. 1, 9 (2020) (“A deepfake could be used to provoke mass panic, by depicting the
President informing citizens of an imminent or ongoing attack on the U.S. A deepfake could
be used to discredit the Supreme Court, by depicting one of the Justices admitting to having
taken bribes.”).
27. See KELLEY M. SAYLER & LAURIE A. HARRIS, CONG. RSCH. SERV., IF11333, DEEP
FAKES AND NATIONAL SECURITY (2021); James Rundle, FBI Warns Deepfakes Might Become
Indistinguishable From Reality, WALL ST. J. (Jan. 17, 2020), https://www.wsj.com
/articles/fbi-warns-deepfakes-might-become-indistinguishable-from-reality-11579257004
[https://perma.cc/7K5H-VLHX].
28. See Westerlund, supra note 7, at 39; Chesney & Citron, supra note 11, at 1772
(“Blackmailers might use deep fakes to extract something of value from people, even those
who might normally have little or nothing to fear in this regard, who (quite reasonably)
doubt their ability to debunk the fakes persuasively, or who fear that any debunking would
fail to reach far and fast enough to prevent or undo the initial damage.”).
29. Jon Bateman, Deepfakes and Synthetic Media in the Financial System: Assessing
Threat Scenarios 15–17 (Carnegie Endowment for Int’l Peace, Cybersecurity and the
Financial System Working Paper No. 7, 2020), https://carnegieendowment.org/files/
Bateman_FinCyber_Deepfakes_final.pdf [https://perma.cc/98EP-QX6K].
30. Chesney & Citron, supra note 11, at 1763.
31. Westerlund, supra note 7, at 44.
8 Columbia Journal of Law & Social Problems [55:1

“MrDeepFakes,” however, have emerged to cater to those

interested in deepfake porn. 32 MrDeepFakes, for example, has a
forum, with thousands of comments and hundreds of thousands of
views, about how to make pornographic deepfakes, and it even has
a forum for requesting such deepfakes to be made. 33 Viewers of
videos on sites like MrDeepFakes almost certainly know they are
fake; nonetheless, these deepfakes are humiliating to those they
target. 34

3. Non-Harmful Deepfakes

Deepfakes are often used in non-confusing ways for improving

communication, the arts, and entertainment. The fakery may
either be disclosed or obviously understood from the context of the
video or the context in which the technology is being used. For
example, the Salvador Dalí museum in St. Petersburg, Florida
used deepfake technology to bring Dalí “back to life” for
educational purposes. 35 Similarly, filmmakers received
permission to use Carrie Fisher’s and Peter Cushing’s personas to
create deepfaked versions of Princess Leia and Governor Tarkin,
respectively, in Rogue One: A Star Wars Story. 36
While deepfake technology is most often used in the arts and
entertainment, it also has been used to alter facial expressions in
order to match translated speech or to help people with
Alzheimer’s recognize their loved ones by making their faces

32. See JOHANNES TAMMEKÄND ET AL., SENTINEL, DEEPFAKES 2020: THE TIPPING POINT
82 (2020), https://thesentinel.ai/media/Deepfakes%202020:%20The%20Tipping%20Point,%
20Sentinel.pdf [https://perma.cc/YAU9-9UUR].
33. Forum Requests, MRDEEPFAKES (2021), https://mrdeepfakes.com/forums/forum-
requests [https://perma.cc/GH6U-45NY] (Stipulating only three limits to the types of
deepfakes that can be requested: “1) All requests should be of celebrities only. 2) Celebrities
all have to be 18+ of age, preferably over 21. 3) You need 3+ posts to create a request.”).
34. As will be further discussed in Part II, torts such as intentional infliction of
emotional distress or the right of publicity are likely best equipped for these situations,
where there is harm to dignity but no confusion. This Note, however, focuses on those
deepfakes, detailed in the paragraphs above, where the harm stems from viewer confusion.
35. Dalí Lives (via Artificial Intelligence), SALVADOR DALÍ MUSEUM,
https://thedali.org/exhibit/dali-lives/# [https://perma.cc/4VDA-7L4W] (the technology
allowed “visitors to interact with an engaging life-like Salvador Dalí on a series of screens
throughout the Museum.”). See also Chesney & Citron, supra note 11, at 1769.
36. See Samit Sarkar, Rogue One Filmmakers Explain How They Digitally Recreated
Two Characters, POLYGON (Dec. 27, 2016), https://www.polygon.com/2016/12/27
/14092060/rogue-one-star-wars-grand-moff-tarkin-princess-leia [https://perma.cc/4KQR-AT
B9]. See also Chesney & Citron, supra note 11, at 1770.
2021] Is This Video Real? 9

appear younger. 37 Synthesia, for example, is a program that

allows users to adapt sound and facial movement in videos to
match dozens of different languages. 38 Some scholars, like Ira
Kemelmacher-Shlizerman, believe deepfake technology can also be
useful for easing bandwidth constraints during video chats: since
audio uses far less bandwidth than video, deepfake technologies
can reduce bandwidth by turning audio transmissions into lip-
synced, realistic-looking video using locally-stored video and image
files to re-construct the faces of people with whom one is video-
chatting. 39
Other uses of deepfakes are parodic, sometimes signposting in
their title or through a watermark that the video is not real. A
YouTube user, “Ctrl Shift Face,” for example, created a deepfake
titled, “The Shining starring Jim Carrey : Episode 1 -
Concentration [DeepFake],” which depicted scenes from The
Shining but with Jack Nicholson’s face replaced by Jim Carrey’s. 40
In addition to the disclosure of fakery in the title, the description
under the video says “it’s a deepfake.” 41 Other parodic examples
are even more obvious from the context. For example, one
deepfake, posted to YouTube by user “TheFakening,” depicts
Donald Trump giving a speech alongside his spouse, Melania
Trump, but with Vladimir Putin’s face superimposed over hers. 42
Even if the video’s title were not “Melania Trump Deepfake with
Vladimir Putin’s Face,” 43 the average viewer would likely
recognize the video as fake.

37. See Westerlund, supra note 7, at 41.

38. See Synthesia STUDIO Features, SYNTHESIA, https://www.synthesia.io/features
[https://perma.cc/RM2A-XBVL]; Malaria Must Die, David Beckham Speaks Nine Languages
to Launch Malaria Must Die Voice Petition, YOUTUBE (Apr. 9, 2019),
https://www.youtube.com/watch?v=QiiSAvKJIHo [https://perma.cc/K42U-72K3] (a
YouTube video using deepfake technology to depict David Beckham speaking in nine
different languages to advocate for Malaria Must Die).
39. See Adrienne LaFrance, The Technology That Will Make It Impossible for You to
Believe What You See, ATLANTIC (July 11, 2017), https://www.theatlantic.com
/technology/archive/2017/07/what-do-you-do-when-you-cannot-believe-your-own-eyes/
533154/ [https://perma.cc/C9TD-877V].
40. See Ctrl Shift Face, The Shining starring Jim Carrey : Episode 1 - Concentration
[DeepFake], YOUTUBE (July, 8, 2019), https://www.youtube.com/watch?v=HG_NZpkttXE
[https://perma.cc/E53K-TKAC].
41. Id.
42. TheFakening, Melania Trump Deepfake with Vladimir Putin’s Face, YOUTUBE
(Mar. 11, 2019), https://www.youtube.com/watch?v=K3kgBStHFHs&list=PLWfNGHKv4
EwH64nYpSxT8qh5JnWGj8ymO&index=1 [https://perma.cc/K9M5-35E7].
43. Id.
10 Columbia Journal of Law & Social Problems [55:1

Deepfakes intended to be parodic may, however, toe the line

between obvious parody and confusion as to association, even if
they are not intended to deceive their viewers. Advertisers, for
example, have used deepfakes, sometimes evocatively, to re-depict
famous actors or news anchors in their younger years. 44 State
Farm and ESPN, for example, aired a commercial featuring
SportsCenter anchor Kenny Mayne digitally manipulated to
appear younger. 45 The ad promoted ESPN’s documentary series,
“The Last Dance,” with Kenny Mayne reporting on the Chicago
Bull’s 1998 NBA championship:

“This is the kind of stuff that ESPN will eventually make a

documentary about,” Mr. Mayne says. “They’ll call it
something like ‘The Last Dance.’ They’ll make it a 10-part
series and release it in the year 2020. It’s going to be lit. You
don’t even know what that means yet.” As a vintage State
Farm logo appears in the background, he adds, “And this clip
will be used to promote the documentary in a State Farm
commercial.” 46

The creators said, “We tried to make the joke clear enough so that
we weren’t tricking anyone.” 47 Depending on one’s media literacy
or experience watching SportsCenter, the advertisement may or
may not have been confusing. 48
Likewise, Space150, a creative agency, made a knock-off Travis
Scott music video by “feeding the rapper’s lyrics into a text
generator model, and creating beats and melodies with other
neural network programs.” 49 The result, called “Jack Park Canny
Dope Man” by “Travis Bott,” is a “pretty on-brand sound (including
the rapper’s autotuned voice), although the lyrics are pretty much

44. See Tiffany Hsu, An ESPN Commercial Hints at Advertising’s Deepfake Future,
N.Y. TIMES (Apr. 22, 2020), https://www.nytimes.com/2020/04/22/business/media/espn-
kenny-mayne-state-farm-commercial.html [https://perma.cc/3FSX-M36P].
45. Id.
46. Id.
47. Id.
48. Media reports on the commercial indicate some viewers were initially confused by
it. Charles Curtis, A Prescient State Farm Ad with Kenny Mayne Had Everyone Fooled
During ‘The Last Dance’, USA TODAY (Apr. 20, 2017), https://ftw.usatoday.com/2020/04/the-
last-dance-espn-kenny-mayne-state-farm-ad [https://perma.cc/J7QR-XZRU].
49. Thom Waite, Listen to an AI-generated ‘Travis Scott’ song, DAZED (Feb. 14, 2020),
https://www.dazeddigital.com/music/article/47949/1/listen-to-an-ai-generated-travis-scott-
song-artificial-intelligence-deepfake [https://perma.cc/LP5G-MCGU].
2021] Is This Video Real? 11

completely nonsensical.” 50 The video includes footage of a

performer appearing to be Travis Scott, and, notably, the YouTube
version of the music video does not disclaim a relationship with
Travis Scott or disclose that the video is a deepfake. 51 Hence, to
uninformed viewers, the music video may seem to be endorsed by
the hip-hop artist himself, who was not involved in the project. 52
The takeaway from this and the Kenny Mayne deepfake is that
deciding whether a video is harmful by way of confusion as to
association may require significant factual investigation. 53
Considering, though, that the examples discussed above are
arguably beneficial to society because their expressive value seems
to outweigh the limited, if any, harm by way of viewer confusion,
the legal mechanism applied to deepfakes should be able to
distinguish harmful from non-harmful uses. 54

C. NON-LEGAL SOLUTIONS TO HARMFUL DEEPFAKES

Technology companies and government authorities have sought

to create deepfake detection tools to combat their deleterious
effects caused by confusion. These tools can operate by identifying
minor inconsistences such as abnormal reflections, face wobbling,
and unnatural mouth movements. 55 They can also be tailored to
individuals such as celebrities and politicians who are at a
heightened risk of being deepfaked. 56 Microsoft, for example, has
released a Video Authenticator Tool, which analyzes subtle
blending boundaries and shading “to provide a percentage chance,
or confidence score, that the media is artificially manipulated.” 57

50. Id.
51. See Travis Bott, Travis Bott - JACK PARK CANNY DOPE MAN, YOUTUBE (Feb.
13, 2020), https://www.youtube.com/watch?v=3UwLhqcZqxc [https://perma.cc/7TNY-
MPH8].
52. Davis Gianatasio, This Agency Used A.I. to Make Travis Bott, a Bogus Travis Scott,
MUSE BY CLIO (Feb. 12, 2020), https://musebycl.io/digital-data/agency-used-ai-make-travis-
bott-bogus-travis-scott [https://perma.cc/7GJD-73J9].
53. See infra Part II.
54. See Brown, supra note 26, at 34 (“Banning deepfakes altogether would not only
stifle these positive uses but would also raise insurmountable First Amendment hurdles.”).
As discussed in Part III, trademark law has a well-developed likelihood of confusion doctrine
and First Amendment defense, which courts can apply to ensure socially beneficial uses,
such as parody, are not stifled.
55. See Westerlund, supra note 7, at 45.
56. See id. at 46.
57. Tom Burt & Eric Horvitz, New Steps to Combat Disinformation, MICROSOFT ON THE
ISSUES (Sept. 1, 2020), https://blogs.microsoft.com/on-the-issues/2020/09/01/disinformation-
deepfakes-newsguard-video-authenticator/ [https://perma.cc/9ALN-NAPS].
12 Columbia Journal of Law & Social Problems [55:1

Likewise, malware and cybersecurity companies have developed

their own deepfake detection tools—Zemana, a cybersecurity
company, created a free program called Deepware, which scans
user-uploaded hyperlinks for deepfakes. 58 Despite the availability
of these tools, they are not foolproof—researchers have
demonstrated the ways such detectors can be tricked. 59
In light of the opportunity for improvement in deepfake
detection technology, some companies have sponsored
competitions to encourage innovation. Facebook, for example,
organized a Deepfake Detection Challenge, which drew over 2,000
participants; notably, the winning model had an accuracy rate of
just 65%. 60 Other technology companies like Google have also
made certain datasets publicly available in order to support
development of detection tools. 61 Government organizations too,
including the Defense Advanced Research Projects Agency
(DARPA) of the U.S. Department of Defense, have launched Media
Forensics taskforces to pursue deepfake detection solutions. 62 In
2020 the House of Representatives passed the Identifying Outputs
of Generative Adversarial Networks (IOGAN) Act, which would, if
enacted into law, direct certain agencies to support research on
deepfake detection tools. 63

58. See DEEPWARE, https://scanner.deepware.ai/result/1a93e77865279476998be27

dd6e0a99d70944f13-1608486470/ [https://perma.cc/CQD5-WKVL].
59. University of California San Diego, Deepfake Detectors Can be Defeated, Computer
Scientists Show for the First Time, SCIENCEDAILY (Feb. 8, 2021), www.sciencedaily.com
/releases/2021/02/210208161927.htm [https://perma.cc/HLN4-QHB3].
60. Cristian C. Ferrer et al., Deepfake Detection Challenge Results: An Open Initiative
to Advance AI, FACEBOOK AI (June 12, 2020), https://ai.facebook.com/blog/deepfake-
detection-challenge-results-an-open-initiative-to-advance-ai/ [https://perma.cc/F34V-CU
ZU] (Facebook launched the challenge by sharing a dataset of more than 100,000 videos,
which participants used as a foundation for developing their detection models.).
61. Nick Dufour & Andrew Gully, Contributing Data to Deepfake Detection Research,
GOOGLE AI BLOG (Sept. 24, 2019), https://ai.googleblog.com/2019/09/contributing-data-to-
deepfake-detection.html [https://perma.cc/8U6V-ELJB].
62. Matt Turek, Media Forensics (MediFor), DEF. ADVANCED RSRCH. PROJECTS
AGENCY, https://www.darpa.mil/program/media-forensics [https://perma.cc/HR3K-L6JF].
63. IOGAN Act, Pub. L. No. 116-258 (2020) (directing the National Science Foundation
and the National Institute of Standards and Technology to support research on generative
adversarial networks). U.S. legislators have sponsored other bills, which have not passed,
calling for the Department of Homeland Security to research and develop deepfake detection
tools. See DEEP FAKES Accountability Act, H.R. 3230, 116th Cong. (as referred to H.
Subcomm. Terrorism & Homeland Sec., June 28, 2019). See also Deepfake Report Act of
2019, S. 2065, 116th Cong. (as referred to H. Subcomm. Consumer Prot. & Com., October
29, 2019); Deepfake Task Force Act, S. 2559, 117th Cong. (as referred to S. Comm. on
Homeland Sec. and Gov. Affairs).
2021] Is This Video Real? 13

In addition to efforts to create deepfake detection tools,

websites like Twitter have asserted the right to flag or remove
videos that “deceptively promote synthetic or manipulated media
that are likely to cause harm.” 64 As another solution, some
scholars have discussed the idea of an “immutable life log or
authentication trails that make it possible for a victim of a deep
fake to produce a certified alibi credibly proving that he or she did
not do or say the thing depicted.” 65 Others have proposed digital-
watermark solutions that verify the authenticity of video
recordings contemporaneously with their production—Canon, for
example, has released an “Original Data Security Kit” for verifying
the originality of videos made using its products. 66
Despite these efforts, technologists and legal scholars remain
concerned that deepfake creators will continue to outsmart current
and emerging detection tools. 67 Hany Farid, a computer scientist
at the University of California at Berkeley, has expressed concern
that those working on detection are “out-gunned” 100 to 1 by those
creating deepfakes. 68 Additionally, even if effective detection tools
are developed, it may be difficult to spread awareness about the
deepfake, or it may be too late—the harm caused by the deepfake
may have already taken place. 69 Moreover, some scholars have

64. HELP CENTER, Synthetic and Manipulated Media Policy, TWITTER,

https://help.twitter.com/en/rules-and-policies/manipulated-media [https://perma.cc/49DN-
W2SL]. See also Monika Bickert, Enforcing Against Manipulated Media, FACEBOOK (Jan.
6, 2020), https://about.fb.com/news/2020/01/enforcing-against-manipulated-media/
[https://perma.cc/J4XD-74Z5].
65. See, e.g., Chesney & Citron, supra note 11, at 1814–17. Professors Chesney and
Citron note, however, that such a solution’s costs to individual privacy may outweigh its
potential benefits. Id. at 1815–17.
66. See Spivak, supra note 8, at 353–54 (but noting that such technologies have been
hacked and “rendered useless”).
67. See Brown, supra note 26, at 25. See also James Vincent, Deepfake Detection
Algorithms Will Never Be Enough, THE VERGE (June 27, 2019),
https://www.theverge.com/2019/6/27/18715235/deepfake-detection-ai-algorithms-accuracy-
will-they-ever-work [https://perma.cc/DV47-5Z5S] (warning that detectors will not be able
to keep up with technological advancements allowing for increasingly realistic deepfakes).
68. Drew Harwell, Top AI Researchers Race to Detect ‘Deepfake’ Videos: ‘We are
Outgunned’, WASH. POST (June 12, 2019), https://www.washingtonpost.com
/technology/2019/06/12/top-ai-researchers-race-detect-deepfake-videos-we-are-outgunned/
[https://perma.cc/MJ93-FT9Z].
69. Even if social media companies have access to deepfake detection tools, it may be
difficult for them to decide when to remove deepfakes from their sites. Although some
deepfakes may be obviously fake or otherwise harmless, the detection tool may not be able
to distinguish these from more pernicious deepfakes. See Brown, supra note 26, at 31.
Further, there is the concern that the algorithms of social media sites may spread
disinformation faster and wider than the truth. Id. at 36 (citing Jared Schroeder,
14 Columbia Journal of Law & Social Problems [55:1

noted the public may not trust detection tools, even those that are
effective. 70 Public distrust in social media companies, for example,
may impede acceptance and adoption of such companies’ deepfake
detection tools. 71
Even if detection tools become effective at removing harmful
deepfakes from public view, such tools fail to disincentivize the
creation of harmful deepfakes ex ante. For these reasons, the only
resort for victims of deepfakes may be ex post, through the courts,
whose adversarial system and power to sanction may provide the
trustworthiness and disincentivizing power needed combat
pernicious deepfakes.

II. LEGAL STRATEGIES FOR ADDRESSING DEEPFAKES

As deepfakes have both pernicious and non-pernicious uses, as

discussed above, any legal approach must be narrowly tailored and
not sweep so broadly as to stifle productive speech. Indeed, a
wholesale ban on deepfakes would violate the First Amendment. 72
This Note argues Section 43(a) of the Lanham Act is well-suited
for the task of addressing deepfakes due to its focus on consumer
confusion, its well-established First Amendment protections, and
the fact the Communications Decency Act does not prohibit false
association claims against website owners. To be clear, this Note
does not purport to argue the Lanham Act is the only appropriate,
or necessarily the best, legal mechanism for combatting pernicious
deepfakes in all circumstances. Rather, it provides reasons why
the Lanham Act is well-suited for the task of taking on deepfake
creators and website owners hosting deepfakes, assesses the

Marketplace Theory in the Age of AI Communicators, 17 FIRST AMEND. L. REV. 22, 30

(2018)).
70. Brown, supra note 26, at 26 (noting an individual’s level of confidence in the
deepfake detection software may depend on factors such as ideological views, political
biases, and the subject matter of the deepfake, among others).
71. A 2019 poll by Pew Research Center indicated a substantial portion of the American
public distrusted social media sites, with 59% of those surveyed distrusting Facebook. See
Mark Jurkowitz & Amy Mitchell, An Oasis of Bipartisanship: Republicans and Democrats
Distrust Social Media sites for Political and Election News, PEW RSCH. CTR. (Jan. 29, 2020),
https://www.journalism.org/2020/01/29/an-oasis-of-bipartisanship-republicans-and-democr
ats-distrust-social-media-sites-for-political-and-election-news/ [https://perma.cc/BS92-H6
QT] (“[B]oth Democrats and Republicans (including independents who lean toward either
party)—in an unusual display of bipartisan convergence—register far more distrust than
trust of social media sites as sources for political and election news. And the most distrusted
are three giants of the social media landscape—Facebook, Instagram and Twitter.”).
72. See Spivak, supra note 8, at 399.
2021] Is This Video Real? 15

merits of other legal strategies that have been proposed, and, in

Part III, articulates the mechanics of Lanham Act false association
claim, which need not be the exclusive cause of action in a
deepfake-related suit.

A. SECTION 43(a)(1)(A) OF THE LANHAM ACT

The false association claim contemplated by this Note is

premised on the theory that the target of a confusing deepfake—
i.e., the video’s human subject, who has an intellectual property
right in the use of her persona—is falsely portrayed to be
associated with or otherwise endorsing the message or actions
depicted in the video. The false association cause of action is
codified in Section 43(a)(1)(A) of the Lanham Act, the United
States trademark and false advertising statute. 73
A principal goal of trademark law and the Lanham Act is
consumer protection—i.e., protecting the public from deception. 74
The text of the Act, its construction by the federal courts, and
congressional statements concurrent to the Act’s passage confirm
consumer protection as a goal of trademark and unfair competition
law. 75 While, historically, the Act focused primarily on fostering
competition and preventing deceptive uses of marks by
competitors, it has gradually expanded to prohibit even
noncompetitive uses of a mark likely to deceive consumers as to
association or endorsement. 76
Section 43(a)(1)(A) of the Lanham Act, in particular, was
initially intended to prohibit false designations of geographic
origin, such as Canadian perfume advertised as “Parfum de

73. See 15 U.S.C. § 1125(a)(1)(A)–(B).

74. See, e.g., 1 J. THOMAS MCCARTHY, MCCARTHY ON TRADEMARKS AND UNFAIR
COMPETITION § 2:22 (5th ed. 2020) (“Trademark protection is consumer protection”); id. at
§ 2:24 (quoting Guthrie Healthcare System v. ContextMedia, Inc., 826 F.3d 27, 50 (2d Cir.
2016) (“An important beneficiary of the trademark system is the public.”)); id. at § 2:1
(tracing the historical origins of trademark and unfair competition law to Anglo-American
Courts of Chancery and Courts of Law, and noting the case law of the latter focused on
protecting against “fraud and deceit practiced by the defendant on the public”).
75. See id.
76. MCCARTHY, supra note 74, at § 5:2 (“The growth of trademark protection from only
competitive parties to include noncompetitive uses of a mark, signaled a large expansion in
the protection of marks. This expansion gradually continued in fits and starts until by the
late twentieth century, it was very clear that a strong mark could be protected against use
on noncompetitive goods or services so long as there was a likelihood of confusion over
sponsorship, affiliation or connection.”).
16 Columbia Journal of Law & Social Problems [55:1

France.” 77 Over time, however, courts interpreted Section

43(a)(1)(A) more broadly to prohibit false designations of source,
sponsorship, and other forms of unfair competition. 78 This more
capacious understanding was codified when Congress amended
Section 43(a) in 1988. 79 Since then, courts have continued to
broadly interpret Section 43(a)(1)(A) as covering a wide range of
rights, including that of an individual to the use of her likeness as
an endorsement. 80 The text of Section 43 that is most relevant to
this Note is reproduced below:

(a) Civil action

(1) Any person who, on or in connection with any goods
or services, or any container for goods, uses in commerce any
word, term, name, symbol, or device, or any combination
thereof, or any false designation of origin, false or misleading
description of fact, or false or misleading representation of
fact, which—
(A) is likely to cause confusion, or to cause mistake, or to
deceive as to the affiliation, connection, or association of
such person with another person, or as to the origin,
sponsorship, or approval of his or her goods, services, or
commercial activities by another person, or
(B) in commercial advertising or promotion,
misrepresents the nature, characteristics, qualities, or
geographic origin of his or her or another person’s goods,
services, or commercial activities, shall be liable in a civil
action by any person who believes that he or she is or is
likely to be damaged by such act. 81

Notably, Section 43(a) does not require registration as a pre-

requisite for suit, and it has “been held to apply to situations that
would not qualify formally as trademark infringement, but that

77. JANE C. GINSBURG ET AL., TRADEMARK AND UNFAIR COMPETITION LAW: CASES AND
Materials 478 (6th ed. 2017).
78. Id.
79. Id.
80. Id. See also MCCARTHY, supra note 74, at § 28:15 (5th ed. 2020) (“[T]he 1989 revised
wording of § 43(a) makes it clearer than ever that § 43(a) is an appropriate vehicle for the
assertion of claims of falsely implying the endorsement of a product or service by a real
person.”).
81. 15 U.S.C. § 1125(a)(1)(A)–(B).
2021] Is This Video Real? 17

involve unfair competitive practices resulting in actual or potential

deception.” 82
Courts have read Section 43(a)(1)(A), specifically, to prohibit
the unauthorized use of an individual’s persona—as manifest in a
video or photograph, 83 an audio-recording of one’s voice, 84 or even,
in some cases, a look-alike robot 85—that “creates the false
impression that the party has endorsed a product or service in
interstate commerce.” 86 Plaintiffs need not show that consumers
are likely to “believe that the owner of the mark actually produced
the item and placed it on the market.” 87 Instead, it is sufficient to
show the public is likely to believe the “mark’s owner sponsored or
otherwise approved” the use of the plaintiff’s image. 88
At issue in Allen v. National Video, for example, was defendant
National Video’s advertisement for its video rental chain, which
starred a Woody Allen look-alike, Phil Boroff, “smiling at the
camera while holding up [National Video’s] V.I.P. card.” 89
Although the look-alike did not explicitly endorse the video rental
chain, the U.S. District Court for the Southern District of New
York granted summary judgment for plaintiff Woody Allen on his

82. Allen v. Nat’l Video, Inc., 610 F. Supp. 612, 625 (S.D.N.Y. 1985) (citing SK & F, Co.
v. Premo Pharm. Lab’ys, Inc., 625 F.2d 1055, 1065 (3d Cir.1980)). See also Cher v. F. Int’l,
Ltd., No. 81 1461 R., 1982 WL 916836, at *6 (C.D. Cal. Jan. 15, 1982) (registration of an
actual trademark is not required for suit under Section 43). See infra Part III.A, discussing
the issue of standing.
83. See, e.g., Downing v. Abercrombie & Fitch, 265 F.3d 994, 1008–09 (9th Cir. 2001)
(reversing summary judgment for defendant on question of false endorsement, where
defendant Abercrombie used professional surfers’ photos without their permission in a
catalog: “A jury could reasonably find that Abercrombie intended to indicate to consumers
that these legendary surfers were endorsing Abercrombie’s merchandise.”); Fifty-Six Hope
Road Music, Ltd. v. A.V.E.L.A., Inc., 778 F.3d 1059, 1067 (9th Cir. 2015) (affirming jury
verdict of false endorsement, where the defendant used an unauthorized image of Bob
Marley on its T-shirts).
84. See, e.g., Waits v. Frito-Lay, Inc. 978 F.2d 1093, 1111 (9th Cir. 1992) (affirming jury
verdict of false endorsement, where defendant Frito-Lay used a voice-impersonator of
plaintiff Tom Waits in its Doritos commercial, which likely caused ordinary consumers to
be “confused as to whether Tim Waits sang on the commercial”).
85. See White v. Samsung Elecs. Am., Inc., 971 F.2d 1395, 1396 (9th Cir. 1992); Wendt
v. Host Int’l, Inc., 125 F.3d 806 (9th Cir. 1997).
86. Allen, 610 F. Supp. at 626. See also MCCARTHY, supra note 74, at § 28:15 (noting
that 43(a)(1)(A) does not require the plaintiff to establish the “traditional indicia” of a
trademark right).
87. Famous Horse Inc. v. 5th Ave. Photo Inc., 624 F.3d 106, 109 (2d Cir. 2010) (quoting
Dallas Cowboys Cheerleaders, Inc. v. Pussycat Cinema, Ltd., 604 F.2d 200, 204 (2d Cir.
1979)).
88. Id. (emphasis in original).
89. Allen v. Nat’l Video, Inc., 610 F. Supp. 612, 617, 627 n.8 (S.D.N.Y. 1985), cited by,
e.g., Est. of Smith v. Cash Money Recs., Inc., No. 14CV2703, 2018 WL 2224993, at *8
(S.D.N.Y. May 15, 2018).
18 Columbia Journal of Law & Social Problems [55:1

false endorsement claim. 90 The court noted, “When a public figure

of Woody Allen’s stature appears in an advertisement, his mere
presence is inescapably . . . interpreted as an endorsement.” 91 To
find false association, the court applied trademark law’s
traditional likelihood of confusion standard—”‘the heart of a
successful claim’ under both the Lanham Act and common law
trademark infringement.” 92
As this Note argued in Part I, harmful deepfakes can generally
be differentiated from non-harmful deepfakes based on whether
they confuse the viewer—where there is no confusion, there is
generally no harm inflicted by the deepfake. 93 Since the “heart of
a successful claim” under the Lanham Act is the likelihood of
confusion test, Section 43(a) of the Lanham Act and its associated
caselaw is equipped to differentiate the good deepfakes from the
bad.
Likewise, trademark law’s well-developed First Amendment
precedent suits it to the task of differentiating between harmful
and non-harmful deepfakes. In Lanham Act disputes, courts
generally apply the Rogers test to balance free speech rights with
protection of consumers from deception in trademark cases. 94 The
two-part Rogers test, discussed further in Part III, upholds
expressive use of a mark unless (i) it has “no artistic relevance” to
the disputed work or, (ii) if there is artistic relevance, the use
“explicitly misleads as to the source or the content of the work.” 95
The test significantly limits the reach of the Lanham Act and
would serve to protect parodic (and non-pernicious) deepfakes from
claims of false association. 96
A significant benefit of a false association action, discussed
further in Part III, is that the Communications Decency Act does
not immunize website owners from federal intellectual property

90. Id. at 627–28, 629–30. Notably, the court stated that in the future the defendant
would be required to either provide a “bold and unequivocal disclaimer” or otherwise “make
it completely clear that he is a look-alike and that the plaintiff has nothing to do” with his
work. Id. at 630 (emphasis in original).
91. Id. at 627 n.8.
92. Allen, 610 F. Supp. at 627 (quoting Standard & Poor’s Corp. v. Commodity Exch.,
Inc., 683 F.2d 704, 708 (2d Cir.1982)).
93. The primary exception to this generalization, discussed supra Part I.B.2, is
deepfake porn, when viewers know the video is fake but there is nevertheless harm to the
target’s dignity.
94. MCCARTHY, supra note 74, at § 28:16.
95. Rogers v. Grimaldi, 875 F.2d 994, 999 (2d Cir. 1989).
96. See Mark A. Lemley, Privacy, Property, and Publicity, 117 MICH. L. REV. 1153, 1173
(2019).
2021] Is This Video Real? 19

claims such as those under Section 43(a)(1)(A) of the Lanham

Act. 97 In fact, Section 230(e)(2) of the CDA explicitly states,
“Nothing in this section shall be construed to limit or expand any
law pertaining to intellectual property.” A plaintiff would still
need to establish the secondary liability of website owners;
however, unlike a defamation claim, a false association claim at
least is not strictly barred by Section 230. 98
Despite these potential advantages of a false association claim
against deepfake creators, the claim has certain downsides. As
will be discussed in Part III, it may be difficult for non-celebrity
plaintiffs to show they have a commercial interest in their likeness,
which some courts require to achieve standing under Section 43(a)
of the Lanham Act. 99 Likewise, the courts may have trouble
accepting that a defendant’s use of a deepfake was “in connection
with any goods or services” for the purposes of Section 43(a). 100
Additionally, as discussed above, one could not bring a Lanham Act
claim against the creators of deepfakes that are harmful to one’s
dignity but unlikely to confuse their viewers, such as pornographic
deepfakes with visible disclaimers. Other downsides include the
lack of standing for a confused consumer (as opposed to the target
of the deepfake) to sue under Lanham Act 43(a), 101 and the
potential difficulty of establishing secondary liability against
website owners, as discussed further in Part III.

B. TORT LAW

While the Lanham Act offers certain advantages for addressing

the issue of deepfakes, other legal strategies are also applicable.
In particular, this Part explores the viability of tort law claims,
including defamation, intentional infliction of emotional distress,
and the privacy torts.
In most jurisdictions, defamation via libel requires a showing of
“(1) a written false and defamatory statement of fact concerning
the plaintiff; (2) that was published by the defendant to a third
party; (3) due to the defendant’s negligence or actual malice,
depending on the status of the person libeled; and (4) special

97. 47 U.S.C. § 230(e)(2).

98. Id.
99. See MCCARTHY, supra note 74, at § 28:17.
100. See 15 U.S.C. § 1125.
101. MCCARTHY, supra note 74, at § 27:39.
20 Columbia Journal of Law & Social Problems [55:1

damages or per se actionability.” 102 Considering the near

impossibility of creating a deepfake by accident, plaintiffs could
generally satisfy the showing of actual malice or reckless disregard
as to the potential falsity of the deepfake, which would be required
if the deepfake involved a public figure. 103 Defamation law’s per se
categories—statements that impute a criminal offense, loathsome
disease, misconduct in business or office, or sexual misconduct—
could also benefit plaintiffs, because claims involving deepfakes in
these per se categories are entitled to general damages, and some
pernicious deepfakes tend to invoke at least one of these
categories. 104 Hence, the defamation tort can potentially provide
civil plaintiffs with robust damages claims. 105
Because many deepfakes are uploaded anonymously, however,
it may be difficult to identify the appropriate defendant, and a
plaintiff’s only remedy may be against a website owner. Hence,
the primary downside to defamation law is that the
Communications Decency Act immunizes websites from liability
for defamation by third-party posters. 106 In other words, a plaintiff
is likely completely foreclosed, under defamation law, from
achieving injunctions against third-party websites onto which a
deepfake was shared. 107 A claim of intentional infliction of
emotional distress, which requires a showing of “extreme and
outrageous conduct” 108 might also be applicable to deepfakes
depicting especially offensive actions; however, the CDA similarly
would bar such a claim against website owners. 109
In addition to defamation and intentional infliction of emotional
distress, the privacy torts—intrusion upon seclusion, publicity
given to private life, false light, and right of publicity
(misappropriation)—are relevant to deepfakes, but each has a
different potential for success. The intrusion upon seclusion tort

102. Spivak, supra note 8, at 370 (quoting Daytree at Cortland Sq., Inc. v. Walsh, No.
15CV2298JFBAYS, 2018 WL 3869247 at *9 (E.D.N.Y. Aug. 15, 2018). See also Celle v.
Filipino Reporter Enters., 209 F.3d 163, 176 (2d Cir., 2000)., cited by Russell Spivak,
“Deepfakes”: The Newest Way to Commit One of the Oldest Crimes, 3 GEO. L. TECH. REV.
339, 366 (2019).
103. Chesney & Citron, supra note 11, at 1793. See also Spivak, supra note 8, at 366
(citing New York Times Co. v. Sullivan, 376 U.S. 254 (1964)).
104. See Spivak, supra note 8, at 374–75.
105. Chesney & Citron, supra note 11, at 1795.
106. 47 U.S.C. § 230(c)(1).
107. See id.
108. See Chesney & Citron, supra note 11, at 1753.
109. See 47 U.S.C. § 230(c)(1).
2021] Is This Video Real? 21

is based on the manner by which an intruder acquires information

rather than on the falsity of information. 110 As deepfakers
generally create videos from publicly available photos or videos of
their targets, they are unlikely to be covered by the intrusion upon
seclusion tort. 111 Similarly, the public disclosure of private fact
tort is generally inapplicable to deepfakes, because one of the tort’s
“essential” elements is that the facts at issue are actually true; 112
the damaging aspects of most deepfakes—often egregious
statements or depictions of violent, sexual, or other activities—are
likely to be false and therefore not actionable under the public
disclosure of private fact tort. The false light tort and the right of
publicity, however, are more applicable to deepfakes. The former
is not recognized by many jurisdictions, 113 because it is perceived
as substantially overlapping with the defamation tort. 114 The
latter, however, has gained traction with the courts and state
legislatures, and it is particularly relevant to deepfakes. 115
Although the right of publicity’s origins lie in state privacy law,
it has been described as an intellectual property right and is
similar to the false association claim proposed by this paper. 116 To
establish a claim, the plaintiff usually must show (i) the defendant
used an aspect of the plaintiff’s identity or persona, without
consent, in such a way that the plaintiff is identifiable from the

110. Spivak, supra note 8, at 379. See also RESTATEMENT (SECOND) OF TORTS § 652B
(AM. L. INST. 1977).
111. See Spivak, supra note 8, at 379.
112. Spivak, supra note 8, at 379 (quoting Tyne ex rel. Tyne v. Time Warner Ent. Co.,
L.P., 204 F. Supp. 2d 1338, 1344 (M.D. Fla. 2002), aff’d sub nom. Tyne v. Time Warner Ent.
Co., L.P., 425 F.3d 1363 (11th Cir. 2005). See also, e.g., Wood v. Hustler Mag., Inc., 736 F.2d
1084, 1090, 1093 (5th Cir. 1984) (applying Texas law and finding defendant Hustler liable
for the false light tort but not the public disclosure of private facts tort, because the
information accompanying Hustler’s disclosure of plaintiff’s photo was not truthful), cited
by Spivak, supra note 8, at 378.
113. See, e.g., Jews For Jesus, Inc. v. Rapp, 997 So. 2d 1098, 1113 (Fla. 2008).
114. Id. (“Based upon our review of the law in Florida and in many other jurisdictions,
we simply cannot ignore the significant and substantial overlap between false light and
defamation.”). See also J. Clark Kelso, False Light Privacy: A Requiem, 32 SANTA CLARA L.
REV. 783, 785 (1992) (reviewing hundreds of false light cases and arguing “there is not even
a single good case in which false light can be clearly identified as adding anything distinctive
to the law.”).
115. See Spivak, supra note 8, at 383–86.
116. 1 J. THOMAS MCCARTHY & ROGER E. SCHECHTER, RIGHTS OF PUBLICITY AND
PRIVACY §§ 1:2, 1:3 (2d ed. 2021) (“The right of publicity is a state-law created intellectual
property right whose infringement is a commercial tort of unfair competition.”). Courts and
commentators have even described Lanham Act false endorsement claims as the “federal
equivalent of the right of publicity.” See ETW Corp. v. Jireh Pub., Inc., 332 F.3d 915, 924
(6th Cir. 2003) (citing Bruce P. Keller, The Right of Publicity: Past, Present, and
Future, 1207 PLI CORP. LAW AND PRAC. HANDBOOK, 159, 170 (2000)).
22 Columbia Journal of Law & Social Problems [55:1

defendant’s use, and (ii) defendant’s use is likely to cause damage

to the commercial value of the plaintiff’s persona. 117 Most states
recognize the right of publicity cause of action either via common
law or via statute. 118 New York, for example, has codified the right
of publicity, calling it the “Right of Privacy,” which creates a
private action for injunction and damages for:

Any person whose name, portrait, picture or voice is used

within this state for advertising purposes or for the purposes
of trade without the written consent first obtained . . . 119

As deepfakes appropriate the persona of others for their own

purposes, the right of publicity is a particularly relevant cause of
action. Although it is similar to the false association cause of
action, it differs in one critical way—falsity or likelihood of
confusion is not an element of the right of publicity tort. 120 Instead,
central to a right of publicity claim is the identifiability of the
plaintiff in the allegedly infringing article. 121
The right of publicity’s focus on identifiability, rather than on
confusion, can lead to different results than false association
claims. In Allen v. National Video, Inc., for example, the court
concluded that plaintiff Woody Allen’s suit against National Video
for its use of a Woody Allen look-alike was “more properly regarded
as one for unfair competition [i.e., false association] under the
Lanham Act,” as opposed to one for the right of publicity. 122 In
Allen, there was “at least a likelihood of consumer confusion as to
whether plaintiff endorse[d] National Video,” making the Lanham
Act well-suited for the claim. 123 Contrastingly, the fact that the
defendant used a look-alike, rather than an image of Woody Allen
himself, presented “difficult questions of law and fact” as to
whether National Video could be liable under a right of publicity
claim, which requires use of the plaintiff’s “portrait” or “picture.” 124

117. MCCARTHY & SCHECHTER, supra note 116, at § 3:2.

118. See id. at §§ 1:2, 1:36, 6:6.
119. N.Y. CIV. RIGHTS LAW § 51 (McKinney 2021).
120. MCCARTHY, supra note 74, at § 28:7.
121. Id. at § 28:12.
122. Allen v. Nat’l Video, Inc., 610 F. Supp. 612, 632 (S.D.N.Y. 1985).
123. Id.
124. Id. at 632, 624 (“The substantive standard of likelihood of confusion provided by
the Lanham Act, discussed below, seems more appropriate than the somewhat strained
construction required here under section 51. . . . Moreover, the relief available to plaintiff is
virtually identical under either theory.”).
2021] Is This Video Real? 23

Hence, the court resolved the matter under the Lanham Act
instead of under the New York right of publicity statute, granting
summary judgment on the false endorsement claim in favor of
Allen. 125 However, in other cases, it may be easier to establish
“identifiability” under a right of publicity theory than confusion
under a false association theory. As McCarthy has noted:

All other things being equal, it will often be easier for a

plaintiff to prove “identifiability” infringement of the right of
publicity than to prove “likelihood of confusion” infringement
of a trademark right. A defendant’s unpermitted use which
does not satisfy the trademark test of “likelihood of
confusion” may satisfy the lesser “identifiability” test of the
right of publicity. As Judge Nies observed: “There may be no
likelihood of such confusion as to the source of goods even
under a theory of ‘sponsorship’ or ‘endorsement,’ and,
nevertheless, one’s right of privacy, or the related right of
publicity, may be violated.” 126

However, as this Note contends that viewer confusion is the

touchstone of most pernicious deepfakes, 127 the Lanham Act’s focus
on confusion makes the law generally well-suited to ensuring only
worthy claims proceed against deepfake creators.
The First Amendment doctrine as applied to false association
claims is similarly focused on confusion, unlike in the case of the
right of publicity. Two cases decided on the same day by the Ninth
Circuit—Brown v. Electronic Arts, Inc. and In re NCAA Student-
Athlete Name & Likeness Licensing Litigation—demonstrate how
courts treat First Amendment defenses to false association and to
the right of publicity differently. 128 At issue in both cases was EA
Sports’ use of athletes’ likenesses in its football video games. The
facts of the cases were essentially identical: in both, EA Sports
used the plaintiffs’ likenesses—their recognizable facial features,
skin tone, hair color, age, height, weight, etc. (but not their actual

125. Id. at 624.

126. MCCARTHY, supra note 74, at § 28:12 (quoting Univ. of Notre Dame Du Lac v. J.C.
Gourmet Food Imps. Co., 703 F.2d 1372, 1376 (Fed. Cir. 1983)).
127. See supra Part II.
128. Brown v. Elec. Arts, Inc., 724 F.3d 1235 (9th Cir. 2013); In re NCAA Student-
Athlete Name & Likeness Licensing Litig., 724 F.3d 1268 (9th Cir. 2013). See also Lemley,
supra note 96, at 1175–77.
24 Columbia Journal of Law & Social Problems [55:1

names)—without consent in their video games. 129 There was,

however, one critical difference between the cases: in Brown, the
plaintiff formulated his claim in terms of false association under
Section 43(a) of the Lanham Act; whereas, in NCAA, the plaintiffs
formulated their claims under California’s common law and
statutory right of publicity. 130 This difference resulted in the
Ninth Circuit rejecting the plaintiff’s false association claim using
the Rogers’ First Amendment test in Brown, but upholding the
plaintiffs’ right of publicity claim in NCAA, where the court did not
apply Rogers, reasoning:

[W]e disagree that the Rogers test should be imported

wholesale for right-of-publicity claims. Our conclusion on
this point is consistent with the Third Circuit’s rejection of
EA’s identical argument in Hart. . . . As the history and
development of the Rogers test makes clear, it was designed
to protect consumers from the risk of consumer confusion—
the hallmark element of a Lanham Act claim. . . . The right
of publicity, on the other hand, does not primarily seek to
prevent consumer confusion. See Hart, 717 F.3d at 158
(“[T]he right of publicity does not implicate the potential for
consumer confusion . . . .”). Rather, it primarily “protects a
form of intellectual property [in one’s person] that society
deems to have some social utility.” . . . The right of publicity
protects the celebrity, not the consumer. Keller’s publicity
claim is not founded on an allegation that consumers are
being illegally misled into believing that he is endorsing EA
or its products.131

Courts have at times, and sometimes inconsistently, applied the

Rogers test to right of publicity claims. 132 In NCAA, the Ninth

129. See Brown, 724 F.3d at 1240; In re NCAA Student-Athlete Name & Likeness
Licensing Litig., 724 F.3d at 1271.
130. See Brown, 724 F.3d at 1240; In re NCAA Student-Athlete Name & Likeness
Licensing Litig., 724 F.3d at 1272.
131. In re NCAA Student-Athlete Name & Likeness Licensing Litig., 724 F.3d at 1280
(internal citations omitted). Jennifer Rothman has criticized this difference in First
Amendment treatment between the right of publicity and false endorsement, arguing the
lack of harmony allows the right of publicity to “thwart the speech protections provided for”
in the Lanham Act. JENNIFER ROTHMAN, THE RIGHT OF PUBLICITY: PRIVACY REIMAGINED
FOR A PUBLIC WORLD 158 (2018).
132. See id. at 1281 (discussing the Sixth Circuit’s inconsistent application of Rogers in
right of publicity cases, and the Eighth and Tenth Circuits’ rejection of the Rogers test in
right of publicity cases); MCCARTHY & SCHECHTER, supra note 116, at § 8:23 (noting that
2021] Is This Video Real? 25

Circuit refused to apply Rogers and instead applied a five factor

transformative use test, which proved less protective of free speech
than the Rogers test. 133 Considering confusion is deepfakes’
principal harm and that there are numerous beneficial uses of
deepfakes, the arguably more protective, confusion-focused Rogers
test, 134 as applied in the context of the Lanham Act, is best suited
to filter the unworthy from the worthy claims against deepfake
creators.
The most significant drawback to the right of publicity claim is
that certain courts do not allow plaintiffs to proceed against
website owners for claims of secondary liability. Although the
Communications Decency Act provides an exception from
immunity for intellectual property claims, and the right of
publicity has been construed as an intellectual property right, the
Ninth Circuit has ruled that the CDA’s exception for intellectual
property only extends to federal (i.e., not state) intellectual
property claims, hence barring right of publicity claims against
website owners on theories of secondary liability. 135 Finally,
another disadvantage to the right of publicity tort, as compared to
false association, is states’ diverse treatment of it. As the right of
publicity differs state by state, 136 injunctions may be limited in
geographical scope. 137 Because the Lanham Act is federal law,

only the Sixth circuit has applied the Rogers test in right of publicity cases and stating that
the problem with applying the Rogers test is it is “a falsity-based test which does not fit with
the right of publicity, which has no requirement of falsity.”).
133. In re NCAA Student-Athlete Name & Likeness Licensing Litig., 724 F.3d at 1274.
134. Contra Rogers v. Grimaldi, 875 F.2d 994, 1004 (2d Cir. 1989) (“Because the right of
publicity, unlike the Lanham Act, has no likelihood of confusion requirement, it is
potentially more expansive than the Lanham Act. Perhaps for that reason, courts
delineating the right of publicity, more frequently than in applying the Lanham Act, have
recognized the need to limit the right to accommodate First Amendment concerns.”)
(citations omitted).
135. Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102, 1119 (9th Cir. 2007) (limiting the
CDA’s intellectual property exception to immunity for federal intellectual property law
claims). But see Hepp v. Facebook, 14 F.4th 204, 206 (3d Cir. 2021) (holding that Section
230 does not bar right of publicity claims, because they fall under Section 230(e)(2)’s
intellectual property exception). See also MCCARTHY & SCHECHTER, supra note 116, at
§ 3:19.
136. See MCCARTHY & SCHECHTER, supra note 116, at § 6:6.
137. See Allen v. Nat’l Video, Inc., 610 F. Supp. 612, 630 (S.D.N.Y. 1985) (“Defendants
have argued that any injunction against them must be limited in geographical scope to New
York State. While such a limitation might be required for an injunction under the New
York Civil Rights Law, given the differences in privacy law among different jurisdictions,
an injunction under the Lanham Act need not be so limited.”).
26 Columbia Journal of Law & Social Problems [55:1

however, injunctions resulting from false association actions are

similarly enforceable across all U.S. jurisdictions. 138
Despite these disadvantages, the right of publicity has two
potential advantages over false association. One such advantage
is the right of publicity does not necessarily require the plaintiff to
be a public figure with a commercial interest in her identity; 139
whereas, as will be discussed in Part III, some courts require
plaintiffs to have a commercial interest in their identity in order to
state a false association claim. 140 The other advantage of the right
of publicity is it provides a remedy to harms caused by damage to
dignity rather than by confusion or falsity. As discussed previously
in Part I, viewers may be aware that individuals in pornographic
deepfakes did not actually engage in the acts depicted—i.e., such
videos may not cause confusion—hence generally precluding
liability under the Lanham Act. 141 Nevertheless, the targets of
these videos suffer harm to their dignity and privacy, which may
be more appropriately addressed by the right of publicity.

C. COPYRIGHT LAW

In addition to the tort-based claims discussed above,

commentators have considered copyright, regulatory, criminal,
and statutory approaches to deepfakes. Although Section 230 of
the Communications Decency Act does not bar contributory
copyright infringement claims, 142 copyright law is unlikely to offer
substantial relief. As an initial matter, the targets of deepfakes
are unlikely to be copyright owners of the images or videos that
form the basis of the disputed deepfake. 143 Only the owners of the
copyright in the underlying images have standing to sue, and, as
Russell Spivak argues, they are unlikely to do so in order to
vindicate deepfake victims or the general public, given the

138. Id.
139. MCCARTHY & SCHECHTER, supra note 116, at § 3:13 (“The clear majority view is
that a non-celebrity can sue for the unpermitted commercial use of his or her identity.”).
140. MCCARTHY, supra note 74, at § 28:17.
141. In some cases, courts have glossed over the likelihood of confusion issue in order to
find Lanham Act liability due to moralistic concerns. See Dallas Cowboys Cheerleaders,
Inc. v. Pussycat Cinema, Ltd., 604 F.2d 200, 205 (2d Cir. 1979) (“Indeed, it is hard to believe
that anyone who had seen defendants’ sexually depraved film could ever thereafter
disassociate it from plaintiff’s cheerleaders.”).
142. See 47 U.S.C. § 230(e)(2) (stating that Section 230 does not limit intellectual
property laws).
143. See Spivak, supra 8, at 396–98.
2021] Is This Video Real? 27

expenses of litigation and the arguably minimal harm they

incurred as a result of the infringement. 144 Further, as Professor
Nina Brown has argued, even if the photos or videos from which a
deepfake is derived are subject to copyright protection, the
resulting deepfake is a new creation, which may be upheld under
copyright’s fair use doctrine. 145 One potential benefit of copyright
law is the notice and takedown process of the Digital Millennium
Copyright Act (DMCA), which provides a less-costly and
potentially more effective alternative to copyright suit. The DMCA
incentivizes websites to take down infringing material by
providing immunity from infringement suits for doing so in
compliance with the act’s procedures. 146 However, as Russell
Spivak notes, the copyright owner rather than the target of the
deepfake would have to initiate the process. 147 Further, the
copyright owner would ultimately still need to have a valid
copyright claim against the alleged infringer, which, as noted, is
uncertain, given copyright’s fair use doctrine. 148

D. REGULATORY SOLUTIONS

Other scholars have addressed the possibility of deepfake

regulation by Federal administrative agencies such as the Federal
Trade Commission (FTC), the Federal Communications
Commission (FCC), or the Federal Election Commission (FEC).
Some deepfakes in the context of advertising may naturally fall
under the FTC’s purview—the regulation of “unfair or deceptive
acts or practices in or affecting commerce.” 149 However, attempts
by the FTC—a government entity—to distinguish between “true”
and “false” information and to regulate false information, is likely
to raise First Amendment concerns. 150 The FCC may be a better
candidate: while the FCC’s jurisdictional authority to regulate

144. See id.

145. See Brown, supra note 26, at 40 n.132; 17 U.S.C.A. § 107 (Westlaw through Pub. L.
No. 117-41).
146. See Spivak, supra note 8, at 398–99.
147. Id.
148. Id.
149. Federal Trade Commission Act, 15 U.S.C. § 45(a)(1); Chesney & Citron, supra note
11, at 1806. See also Elizabeth Caldera, “Reject the Evidence of Your Eyes and Ears”:
Deepfakes and the Law of Virtual Replicants, 50 SETON HALL L. REV. 177, 178 (2019)
(discussing possible administrative agency solutions to the issue of deepfakes and
ultimately proposing that the FTC is best suited to regulate deepfakes).
150. Chesney & Citron, supra note 11, at 1806.
28 Columbia Journal of Law & Social Problems [55:1

information circulating over social media is debatable, 151 the

agency has asserted its authority to engage in rulemaking on
Section 230 of the CDA. 152 Whether and how, specifically, the FCC
would regulate social media remains to be seen, however.
Professors Chesney and Citron have also discussed the FEC as a
possible regulator of deepfakes, noting the potential for it to
require deepfake creators to disclose fakery in the context of
campaigns. 153 They note, however, that the FEC’s jurisdiction over
deepfakes is dubious: it does not have clear jurisdiction over the
social media platforms on which deepfakes most frequently spread,
it does not currently regulate the truth of campaign statements,
and, regardless, it does not have jurisdiction outside the context of
elections. 154 Moreover, as the targets of deepfakes have the
incentive to sue under defamation, privacy tort, or false association
causes of action, there is less of a rationale for intervention by
government regulators. 155 Most importantly, “the prospect of a
government entity attempting to distinguish real news from fake
news—and suppressing the latter—raises serious First
Amendment concerns.” 156 Together, these concerns pose
significant hurdles to a regulatory solution for deepfakes.

E. STATE AND FEDERAL STATUTES

Criminal law, too, has been invoked as a method of combatting

deepfakes. Texas, for example, was the first state to pass a law
criminalizing deepfakes that are made “with intent to influence
the outcome of an election. 157 The Texas law reads:

151. Id. at 1807.

152. Thomas M. Johnson, Jr. The FCC’s Authority to Interpret Section 230 of the
Communications Act, FEDERAL COMMUNICATIONS COMMISSION (Oct. 21, 2020),
https://www.fcc.gov/news-events/blog/2020/10/21/fccs-authority-interpret-section-230-
communications-act [https://perma.cc/W4PY-B7HW]. See also Darrell M. West, What to
Expect from a Biden FCC on Section 230, Net Neutrality, and 5G, BROOKINGS (Dec. 3, 2020),
https://www.brookings.edu/blog/techtank/2020/12/03/what-to-expect-from-a-biden-fcc-on-
section-230-net-neutrality-and-5g/ [https://perma.cc/UL7F-K3EZ].
153. Chesney & Citron, supra note 11, at 1806.
154. Id.
155. Id.
156. Id.
157. TEX. ELEC. CODE ANN. § 255.004 (West 2019); S.B. 751, 2019 Leg., 86th Sess. (Tex.
2019); Brown, supra note 26, at 9 (arguing the law may not survive strict scrutiny under
the First Amendment, because it prohibits speech based on whether it is false, which has
been rejected by the Supreme Court) (citing United States v. Alvarez, 567 U.S. 709, 718
(2012)).
2021] Is This Video Real? 29

(d) A person commits an offense if the person, with intent to

injure a candidate or influence the result of an election:
(1) creates a deep fake video; and
(2) causes the deep fake video to be published or
distributed within 30 days of an election.
(e) In this section, “deep fake video” means a video, created
with the intent to deceive, that appears to depict a real person
performing an action that did not occur in reality. 158

Likewise, Virginia has passed a law criminalizing the distribution

of nonconsensual deepfake porn as a Class 1 misdemeanor,
punishable by a fine of $2,500 and up to one year in jail. 159 In
addition to passing criminal laws, other states such as California
have passed laws providing for private causes of action against the
creators of deepfakes used for political manipulation and
pornography without permission. 160
Members of U.S. Congress have also proposed legislation
combatting deepfakes. Senator Ben Sasse of Nebraska, for
example, sponsored the Malicious Deep Fake Prohibition Act of
2018, which would make it a federal felony to create deepfakes
intended to “facilitate criminal or tortious conduct under Federal,
State, local, or Tribal law.” 161 That is, the bill would just
criminalize or further criminalize acts that are already prohibited
by law. 162 More recently, Representative Yvette Clarke of New
York proposed the “Defending Each and Every Person from False
Appearances by Keeping Exploitation Subject to Accountability
Act of 2019” (the DEEP FAKES Accountability Act of 2019), which
would require deepfakers to watermark their videos with

158. TEX. ELEC. CODE ANN. § 255.004 (West 2019).

159. VA. CODE ANN. § 18.2-386.2(A) (West 2019) VA. CODE ANN. § 18.2-11(a) (West
2019).
160. CAL. CIV. CODE § 1708.85(b) (West 2020) (providing a private cause of action
against persons who create sexually explicit deepfakes); CAL. ELEC. CODE § 20010(a) (West
2020) (prohibiting distribution of deceptive media “with actual malice” within sixty days of
an election); K.C. Halm et al., Two New California Laws Tackle Deepfake Videos in Politics
and Porn, DAVIS WRIGHT TREMAINE LLP (Oct. 14, 2019), https://www.dwt.com/
insights/2019/10/california-deepfakes-law [https://perma.cc/CCK6-WSQH]. New York, too,
has passed a law providing for a private cause of action against the creators of
nonconsensual pornographic images created through what the New York legislature called
“digitization,” i.e., “deepfakery.” N.Y. CIV. RIGHTS LAW § 52-c (McKinney 2021)
(“‘[D]igitization’ means to realistically depict the nude body parts of another human being
as the nude body parts of the depicted individual”).
161. Malicious Deep Fake Prohibition Act, S. 3805, 115th Cong. (2018).
162. See Brown, supra note 26, at 49.
30 Columbia Journal of Law & Social Problems [55:1

disclaimers identifying the fakery. 163 The bill would establish

criminal penalties, punishable by up to five years in prison, for
failure to include disclaimers when there is also intent to cause
various harms. 164 Additionally the act would provide for civil
penalties of up to $150,000 and would establish a standalone
private right of action. 165
Although these statutes are steps in the right direction, some
of them are rather blunt instruments that inadequately
differentiate between harmful and non-harmful deepfakes. The
Texas statute, for example, provides no exceptions for parodies or
other uses that would be protected by the First Amendment. 166
Instead, the statute vaguely requires that the deepfake be created
“with the intent to deceive” and “with intent to injure a candidate
or influence the result of an election.” 167 U.S. Representative
Yvette Clarke’s proposed DEEP FAKES Accountability Act of 2019
is slightly more nuanced: although its disclaimer requirement
could be construed as a form of compelled speech, 168 it provides
certain exceptions to the disclosure and watermark requirements
to account for parodies and other media where “a reasonable
person would not mistake the falsified material activity for actual
material activity of the exhibited living person.” 169 It does not
clarify, however, who bears the burden of establishing that the
exception applies, which could lead to either over-disclosure or to
stifling parodic free speech by those who do not want to include a
disclaimer. 170 More significantly, due to Section 230 of the
Communications Decency Act, website owners are likely to remain
immunized from secondary liability stemming from the civil causes
of action created by these statutes (unless Section 230 is amended),
thus hampering their effectiveness. 171

163. DEEP FAKES Accountability Act, H.R. 3230, 116th Cong. (2019) (as referred to H.
Subcomm. Terrorism & Homeland Sec., June 28, 2019).
164. Id.
165. Id.
166. TEX. ELEC. CODE ANN. § 255.004 (West 2019).
167. Id.
168. See, e.g., Wooley v. Maynard, 430 U.S. 705 (1977) (finding requirement that New
Hampshire license plates bear the state’s motto “Live Free or Die” constitutes compelled
speech in violation of the First Amendment).
169. DEEP FAKES Accountability Act, H.R. 3230, 116th Cong. (2019) (as referred to H.
Subcomm. Terrorism & Homeland Sec., June 28, 2019).
170. Hayley Tsukayama et al., Congress Should Not Rush to Regulate Deepfakes, ELEC.
FRONTIER FOUND. (June 24, 2019), https://www.eff.org/deeplinks/2019/06/congress-should-
not-rush-regulate-deepfakes [https://perma.cc/A2D9-G6RS].
171. See 47 U.S.C. § 230.
2021] Is This Video Real? 31

III. MECHANICS OF A FALSE ASSOCIATION CLAIM

This Part outlines the key elements of a false association claim

against a deepfake creator or a website hosting deepfakes:
establishing standing, proving the four elements of false
association, addressing a First Amendment defense, and
establishing secondary liability.

A. STANDING

While there is no requirement that a plaintiff show “use in

commerce” of her persona to establish standing under Section 43(a)
of the Lanham Act, 172 some courts have required plaintiffs to
establish a “commercial interest” in their persona to achieve
standing. 173
In 2014, the Supreme Court held in Lexmark Int’l, Inc. v. Static
Control Components, Inc. that a plaintiff must establish a
“commercial interest in reputation or sales” to come within the
“zone of interests” of Section 43(a) to assert a false advertising
claim. 174 Additionally, Lexmark required false advertising
plaintiffs to show “economic or reputational injury flowing directly
from the deception wrought by the defendant’s advertising . . .
[which] occurs when deception of consumers causes them to
withhold trade from the plaintiff.” 175 Courts are split on whether
Lexmark—which concerned a Section 43(a)(1)(B) false advertising
claim—also applies to Section 43(a)(1)(A) false association claims:
some courts, without explanation, have extended Lexmark to false
association, 176 while others have limited Lexmark to its context—

172. See, e.g., Belmora LLC v. Bayer Consumer Care AG, 819 F.3d 697 (4th Cir. 2016).
173. MCCARTHY, supra note 74, at § 28:17.
174. Lexmark Int’l, Inc. v. Static Control Components, Inc, 572 U.S. 118, 131–32 (2014).
175. Id. at 133–34.
176. See, e.g., Lindley v. Puccino’s, Inc., No. CV 19-11414, 2019 WL 4673173, at *2 (E.D.
La. Sept. 25, 2019) (“Applying those principles [of Lexmark] to
Lindley’s false endorsement claim under § 1125(a), Lindley has demonstrated that she is
within the class of plaintiffs authorized to sue under § 1125(a).”); One Source Env’t., LLC v.
M + W Zander, Inc, No. 2:12-CV-145, 2015 WL 7428572, at *11 (D. Vt. Nov. 20, 2015)
(granting summary judgment to defendant on issue of false association because plaintiff
failed to show sufficient “injury to a commercial interest in its sales or reputation”);
Walkowicz v. American Girl Brands, LLC, No. 20-CV-374-JDP, 2021 WL 510729, at *3
(W.D. Wis. Feb. 11, 2021) (applying Lexmark’s commercial-interest requirement to Lanham
Act false-endorsement claim). The U.S. District Court for the Northern District of Illinois
has inconsistently applied Lexmark to false association claims. Compare Martin v. Wendy’s
Int’l, Inc., 183 F. Supp. 3d 925, 932 (N.D. Ill. 2016) (applying Lexmark commercial-interest
requirement to Lanham Act false-endorsement claim), and Yeager v. Innovus Pharm., Inc.,
32 Columbia Journal of Law & Social Problems [55:1

false advertising. 177 One court, which did not apply Lexmark to
false association claims, reasoned there is grounding in the opinion
of Lexmark to limit its standing requirements to false advertising
claims. 178 As the U.S. District Court for the District of Maryland
noted in Hershey Co. v. Friends of Steve Hershey, 179 the Supreme
Court in Lexmark contrasted false association from false
advertising claims, thus appearing to limit its standing
requirement to false advertising claims:

Most of the enumerated purposes [in the Lanham Act,

including that of preventing fraud and deception] are
relevant to false-association cases; a typical false-advertising
case [however] will implicate only the Act’s goal of
“protect[ing] persons engaged in [commerce within the
control of Congress] against unfair competition.” Although
“unfair competition” was a “plastic” concept at common law,
it was understood to be concerned with injuries to business
reputation and present and future sales.
We thus hold that to come within the zone of interests in a
suit for false advertising under § 1125(a), a plaintiff must
allege an injury to a commercial interest in reputation or
sales. 180

No. 18-CV-397, 2019 WL 447743, at *7 (N.D. Ill. Feb. 5, 2019) (same), with Ratchford v.
AEG Ventures, LLC, No. 17 C 7368, 2019 WL 10248701, at *4 (N.D. Ill. Mar. 11, 2019)
(more recently applying Lexmark to 1125(a)(1)(B) false advertising claims but not to
1125(a)(1)(A) false association claims). The D.C. Circuit has cited Lexmark for the
proposition that, to establish the elements of a false association claim (rather than to
establish the threshold issue of standing), the plaintiff must prove a commercial interest
that could be harmed by infringement, which the U.S. District Court for the District of
Columbia interpreted as extending Lexmark to 43(a)(1)(A). See Abrahams v. Simplify
Compliance, LLC, No. CV 19-3009 (RDM), 2021 WL 1197732, at *4 (D.D.C. Mar. 30, 2021)
(citing Paleteria La Michoacana, Inc. v. Productos Lacteos Tocumbo S.A. De C.V., 743 F.
App’x 457 (D.C. Cir. 2018)).
177. See, e.g., Avalos v. IAC/Interactivecorp., No. 13-CV-8351(JMF), 2014 WL 5493242,
at *4 (S.D.N.Y. Oct. 30, 2014) (applying Lexmark to 1125(a)(1)(B) false advertising claims
but not to 1125(a)(1)(A) false association claims); Mitchell v. Sanchez, No. 14-0996-CV-ODS,
2015 WL 1393266, at *2 (W.D. Mo. Mar. 25, 2015); Krupa v. Platinum Plus, LLC, No. 8:16-
CV-3189-T-33MAP, 2017 WL 1050222, at *5 (M.D. Fla. Mar. 20, 2017); Hershey Co. v.
Friends of Steve Hershey, No. WDQ-14-1825, 2015 WL 795841, at *5 (D. Md. Feb. 24, 2015);
Gray v. LG&M Holdings LLC, No. CV-18-02543-PHX-SRB, 2020 WL 6200165, at *10 (D.
Ariz. Sept. 23, 2020) (applying Lexmark to false advertising claim but not to false
association claim); Mitcheson v. El Antro LLC, No. CV-19-01598-PHX-GMS, 2020 WL
7075239, at *15 (D. Ariz. Dec. 3, 2020).
178. Hershey Co., 2015 WL 795841, at *5.
179. Id.
180. Lexmark Int’l, Inc., 572 U.S. at 131–32 (citations omitted).
2021] Is This Video Real? 33

Moreover, imposing heightened standing requirements for false

advertising, relative to false association, is consistent with what
courts did before Lexmark: the Ninth Circuit, for example,
historically required false advertising plaintiffs to be direct
competitors of defendants but did not impose such a requirement
on false association plaintiffs. 181
Despite the argument Lexmark should not extend to false
association, it is important to note that even before Lexmark was
decided in 2014, some courts, including the Ninth Circuit, 182
required plaintiffs to have “at minimum, a present intent to
commercialize a mark [i.e., a persona]” 183 and not just a “mere
potential of commercial interest” 184 to achieve Section 43(a)(1)(A)
standing.
The import of such a commercial interest requirement, whether
pre- or post-Lexmark, is that it may forbid private persons—
individuals who are not public figures or who have insufficient
community recognition—from suing under a theory of false
association. 185 In Stayart v. Yahoo! Inc., for example, the Seventh
Circuit affirmed the denial of standing to a plaintiff who alleged
her name had commercial value due to her general involvement in

181. In Jack Russell Terrier Network of Northern California v. American Kennel Club,
Inc., now superseded by Lexmark with respect to its false advertising holding, the Ninth
Circuit explained its different standing requirements for false advertising and false
association:
We have said that different causes of action alleged pursuant to the different
subsections of 15 U.S.C. § 1125(a) have different standing requirements. Smith
and Waits involved the “false association” prong of § 43 of the Lanham Act, 15
U.S.C. § 1125(a)(1)(A), under which a plaintiff need only allege commercial injury
based upon the deceptive use of a trademark or its equivalent to satisfy standing
requirements. That the parties be competitors “in the traditional sense” is not
required. On the other hand, for standing pursuant to the “false advertising”
prong of § 43(a) of the Lanham Act, 15 U.S.C. § 1125(a)(1)(B), a plaintiff must
show: (1) a commercial injury based upon a misrepresentation about a product;
and (2) that the injury is “competitive,” or harmful to the plaintiff’s ability to
compete with the defendant.
407 F.3d 1027, 1037 (9th Cir. 2005) (citations omitted).
182. Id.
183. Condit v. Star Editorial, Inc., 259 F. Supp. 2d 1046, 1052 (E.D. Cal. 2003) (rejecting
plaintiff’s false endorsement claim, because the plaintiff was a “private figure,” and the
complaint did not allege “(1) a present intent to use her identity for commercial purposes;
(2) any past or existing commercial use of her image and identity, or that it has commercial
value; or (3) that she competes commercially in the use of her persona; and (4) has suffered
commercial injury to such an interest in her image and identity.”). See also Dovenmuehle
v. Gilldorn Mortg. Midwest Corp., 871 F.2d 697, 700–01 (7th Cir. 1989).
184. Hutchinson v. Pfeil, 211 F.3d 515, 521 (10th Cir. 2000).
185. See Condit, 259 F. Supp.2d at 1052. The requirement of a commercial interest also
precludes a consumer from suing for being misled. Lexmark Int’l, Inc., 572 U.S. at 131–32.
34 Columbia Journal of Law & Social Problems [55:1

the community, because there was no evidence she was engaged in

or anticipated engaging in “the commercial marketing of her
identity.” 186 Likewise, in Cohen v. Facebook, Inc., a federal district
court denied standing, for lack of commercial interest, to non-
celebrity Facebook users whose profile photos and names were
used by Facebook to promote its “Friend Finder” feature. 187
Even under the “commercial interest” requirement, however,
many individuals targeted by deepfakes have a strong argument
for achieving standing under Section 43(a)(1)(A). Lexmark is not
insurmountable—the Supreme Court clarified in its decision that
the plaintiff need not be a competitor of the defendant, and that
disparaging remarks can be sufficient to establish proximate
causation sufficient for standing. 188 In some cases involving
deepfakes, standing will be obvious: professional athletes or
celebrities who regularly license their identities to market
products can easily establish a commercial interest. 189 Similarly,
politicians targeted by deepfakes could argue they have a
commercial interest in directing funds to their not-for-profit
campaigns, an interest that could be subverted by a confusing

186. Stayart v. Yahoo! Inc., 651 F. Supp. 2d 873, 881 (E.D. Wis. 2009), aff’d, 623 F.3d
436 (7th Cir. 2010) (denying standing to plaintiff whose false endorsement suit alleged that
Yahoo! users would believe she was associated with or endorsed unsavory results stemming
from a Yahoo! search of her name, “Bev Stayart.”). See also Nieman v. Versuslaw, Inc., No.
12-3104, 2012 WL 3201931, at *5 (C.D. Ill. Aug. 3, 2012), aff’d, 512 F. App’x 635 (7th Cir.
2013) (denying standing to plaintiff, a private person and insurance claims adjuster, who
similarly based his theory of false endorsement on search results for his name on legal
search websites like Versuslaw and Lexis/Nexis).
187. Cohen v. Facebook, Inc., 798 F. Supp. 2d 1090, 1097–98 (N.D. Cal. 2011). Notably,
in both Stayart and Cohen, the plaintiffs were unlikely to establish a likelihood of confusion,
even if they did show a “commercial interest” in their identities. In Stayart, the court
actually found there was no likelihood of confusion: “Even if Stayart is engaged in
commercial activity with respect to her identity, a commonsense reading of the complaint
demonstrates that there could be no likelihood of confusion.” Stayart, 651 F. Supp. 2d at
883. In Cohen, the court made no such finding, but the facts were similarly disfavorable for
a likelihood of confusion: Facebook’s promotion of its Friend Finder service merely indicated
certain Facebook members had used the service; the promotion did not indicate the users
endorsed or affirmatively chose to associate with Facebook’s messaging about the Friend
Finder service. See generally Cohen, 798 F. Supp. 2d, at 1095–98.
188. Lexmark Int’l, Inc. v. Static Control Components, Inc, 572 U.S. 118, 138 (2014)
(“When a defendant harms a plaintiff’s reputation by casting aspersions on its business, the
plaintiff’s injury flows directly from the audience’s belief in the disparaging statements.”).
189. See, e.g., Kournikova v. Gen. Media Commc’ns Inc., 278 F. Supp. 2d 1111, 1120
(C.D. Cal. 2003) (granting Section 43(a) standing to celebrity tennis player). Although it
will be easier for celebrities to show a commercial interest in their identities, even courts
requiring a commercial interest have held Section 43(a) “does not have a requirement that
a plaintiff is a celebrity.” Arnold v. Treadwell, 642 F. Supp. 2d 723, 735 (E.D. Mich. 2009)
(plaintiff had commercial interest where she “introduced evidence showing a present intent
to commercialize her identity” and “offered evidence that she has sought modeling jobs.”).
2021] Is This Video Real? 35

deepfake. Indeed, courts have held that the use of a persona for
charitable endeavors constitutes a commercial interest. 190
Further, even courts purporting to impose a commercial interest
requirement have sometimes allowed non-celebrities to proceed
with their claims. 191 Hence, a commercial interest requirement,
though an obstacle, need not impede false association suits by
deepfake victims, celebrity or not.

B. ASSERTING A PRIMA FACIE CLAIM OF FALSE ASSOCIATION

To prevail on a false association claim, a plaintiff must establish

that the defendant, (i) in commerce, (ii) in connection with goods
or services, (iii) made a false or misleading representation of fact
(iv) that is likely to cause consumer confusion as to the origin,
sponsorship, or approval of goods or services. 192 Of these four
elements of infringement, the second—use in connection with
goods or services—is likely to be the most difficult to establish in
the context of deepfakes.

190. Cairns v. Franklin Mint Co., 24 F. Supp. 2d 1013, 1033 (C.D. Cal. 1998) (For the
purposes of establishing “commercial interest” for Section 43(a) standing, it was
“immaterial” that the plaintiff used profits for charitable purposes; the plaintiff—a deceased
celebrity’s estate and charitable fund—had a “commercial interest” for the purposes of
Section 43(a), because defendant’s use of the celebrity’s persona would confuse consumers
as to whether the plaintiff endorsed the defendant’s products.).
191. See Doe v. Friendfinder Network, Inc., 540 F. Supp. 2d 288, 306 n.20 (D.N.H. 2008)
(denying defendant’s motion to dismiss false association claim, where the plaintiff was a
non-celebrity who did not allege she commercialized or intended to commercialize her
persona, reasoning that damage to commercial value can be presumed when the defendant
uses the plaintiff’s persona in an advertisement without permission). But see Stayart v.
Yahoo! Inc., 651 F. Supp. 2d 873, 881 (E.D. Wis. 2009), aff’d, 623 F.3d 436 (7th Cir. 2010)
(distinguishing Friendfinder Network on the grounds that it did not explicitly discuss the
issue of prudential standing, reasoning that a court is not obliged to raise questions of
prudential standing).
192. Burck v. Mars, Inc., 571 F. Supp. 2d 446, 455 (S.D.N.Y. 2008), cited by, e.g., Toth
v. 59 Murray Enterprises, Inc., No. 15 CIV. 8028 (NRB), 2019 WL 95564, at *5 (S.D.N.Y.
Jan. 3, 2019), aff’d in part sub nom. Electra v. 59 Murray Enterprises, Inc., 987 F.3d 233
(2d Cir. 2021)). See also 15 U.S.C. § 1125(a)(1)(A).
36 Columbia Journal of Law & Social Problems [55:1

1. Use in Commerce

As a threshold matter, it is crucial to note the “use in commerce”

and “in connection with any goods or services” elements are
distinct, 193 although courts have at times conflated the two. 194
“Use in commerce” is a jurisdictional term referring broadly to
Congress’ intent to legislate using its Commerce Power, while “in
connection with any goods or services” limits the types of uses that
can constitute actionable infringement. 195
Courts have interpreted the “use in commerce” requirement of
Section 43(a) “to refer to a use that falls within Congress’s
commerce power.” 196 Although the Lanham Act itself defines “use
in commerce,” courts have generally not applied this definition in
the context of infringement. As McCarthy explains:

Under any of the three bases for recovery (registered or

unregistered infringement and dilution), the Lanham Act
requires that the plaintiff prove that the accused party has
made a “use in commerce” of the accused mark. It has been
argued that the Lanham Act § 45 definition of “use in
commerce” is in effect a statutory requirement that the
accused use be a “trademark” use. Lanham Act § 45 defines

193. See, e.g., United We Stand Am., Inc. v. United We Stand, Am. N.Y., Inc., 128 F.3d
86, 89 (2d Cir. 1997) (analyzing “use in commerce” and “in connection with any goods or
services” separately); Radiance Found., Inc. v. NAACP, 786 F.3d 316, 322 (4th Cir. 2015)
(“An actionable trademark claim does not simply require that the alleged infringer used in
commerce the mark that the trademark holder possesses. It also requires that the
infringer’s use be ‘in connection with’ goods or services in a manner that is ‘likely to cause
confusion’. . . .”); Utah Lighthouse Ministry v. Found. for Apologetic Info. & Rsch., 527 F.3d
1045, 1054 (10th Cir. 2008) (“It is important to distinguish between the merely jurisdictional
‘in commerce’ requirement, see 15 U.S.C. § 1127, and the ‘in connection with any goods and
services’ requirement that establishes a violation of section 43 of the Lanham Act.”).
194. Bosley Med. Inst., Inc. v. Kremer, 403 F.3d 672, 677 (9th Cir. 2005), aff’d in part,
rev’d in part, 403 F.3d 672 (9th Cir. 2005). In that case, the court stated:
The district court ruled that Kremer’s use of Bosley’s mark was noncommercial.
To reach that conclusion, the court focused on the “use in commerce” language
rather than the “use in connection with the sale of goods” clause. This approach
is erroneous. “Use in commerce” is simply a jurisdictional predicate to any law
passed by Congress under the Commerce Clause. . . . Therefore, the district court
should have determined instead whether Kremer’s use was “in connection with a
sale of goods or services” rather than a “use in commerce.”
Id.
195. See Burck, 571 F. Supp. at 454; 15 U.S.C. § 1125(a)(1)(A).
196. Rescuecom Corp. v. Google Inc., 562 F.3d 123, 138 (2d Cir. 2009). See also
Lens.com, Inc. v. 1-800 Contacts, Inc., 686 F.3d 1376, 1380 (Fed. Cir. 2012) (interpreting
“use in commerce” as a jurisdictional predicate indicating Congress’s intent to legislate to
the limits of its Commerce Clause authority).
2021] Is This Video Real? 37

a “use in commerce” as being limited either to certain kinds

of use on goods or use in the sale or advertising of services.
In my view this is not a reasonable reading of the statute.
The Lanham Act § 45 definition of “use in commerce” is
limited to certain defined kinds of use such as on “displays
associated” with the goods or on “tags or labels.” It was
clearly drafted to define the types of “use” that are needed to
qualify a mark for federal registration—not as a candidate
for infringement. It defines the kinds of “use” needed to
acquire registerable trademark rights—not to infringe
them. . . .
[T]he Second, Ninth and Federal Circuits agree with the
author’s position as described above: the Lanham Act § 45
definition of “use in commerce” defines the kind of use needed
to register a mark, not to infringe it. 197

Hence, most courts have held plaintiffs are “not required to

demonstrate that a defendant made use of the mark in any
particular way to satisfy the ‘use in commerce’ requirement.” 198
In Rescuecom Corp. v. Google Inc., the Second Circuit
prominently employed this broad construction of “use in
commerce.” 199 At issue in Rescuecom was Google’s AdWords
program, which lets advertisers pay for their websites to appear as
search results in response to searches for specific keywords. 200 In
the case of Rescuecom, Google’s Keyword Suggestion Tool had
recommended the keyword (and plaintiff’s trademark),
RESCUECOM, to Rescuecom’s competitors, who purchased the
keyword so their advertisements would appear when users

197. MCCARTHY, supra note 74, at § 23:11.50 (5th ed. 2021) (citing Rescuecom Corp., 562
F.3d at 139; Bosley Med. Inst., Inc., 403 F.3d at 677; VersaTop Support Systems, LLC v.
Georgia Expo, Inc., 921 F.3d 1364, 1370 (Fed. Cir. 2019)). The Sixth Circuit has more
narrowly interpreted the “use in commerce” provision to require “trademark use.” Id.
(citing, e.g., Interactive Products Corp. v. a2z Mobile Office Solutions, Inc., 326 F.3d 687,
695 (6th Cir. 2003) (“If defendants are only using [plaintiff’s] trademark in a ‘non-
trademark’ way—that is, in a way that does not identify the source of a product—then
trademark infringement and false designation of origin laws do not apply.”). One rationale
for the Sixth Circuit’s interpretation, noted by McCarthy, is that non-trademark use of a
mark is less likely to trigger a likelihood of confusion. Id.
198. Kelly-Brown v. Winfrey, 717 F.3d 295, 305 (2d Cir. 2013).
199. Rescuecom Corp., 562 F.3d at 124.
200. Id. at 125–27. Google’s AdWords program is now called “Google Ads.” Google
AdWords is now Google Ads, GOOGLE ADS HELP (2021), https://support.google.com/google-
ads/answer/9028765 [https://perma.cc/XD2B-PM8F].
38 Columbia Journal of Law & Social Problems [55:1

searched for “Rescuecom.” 201 The plaintiff, Rescuecom, based its

infringement claim on its allegation that Google users would be
misled to believe that the advertisements appearing in response to
searches for “Rescuecom” were affiliated with or endorsed by the
plaintiff. 202 The court held defendant Google’s sale of keywords to
advertisers through its AdWords program constituted a “use in
commerce,” even though the keyword (plaintiff’s mark,
RESCUECOM) in dispute was not actually displayed to users of
Google in connection with the competitors’ links. 203 This contrasts
with the case of deepfakes, where the plaintiffs’ marks, their
personas, are displayed prominently in the infringing media—the
deepfake videos themselves.
Generally, plaintiffs in deepfake suits will have a strong
argument that deepfakes fall under Congress’ Commerce Clause
authority: deepfakes are shared nationally on websites like
YouTube and Facebook, and widely disseminated videos of public
figures or political leaders are likely to incite responses
unconstrained by state lines. 204 Even if a deepfake is not widely
distributed across state lines, deepfakes targeting political figures,
for example, may affect their target’s ability to raise campaign
funds or to market their identities for other purposes. 205 In
Planned Parenthood Federation of America, Inc. v. Bucci, for
example, the Second Circuit held that defendant Bucci’s use of the
“plannedparenthood.com” domain name met the “use in
commerce” requirement even if defendant’s activities alone did not

201. Id. Although the court held the use in commerce element was established, it was
unlikely the plaintiff could establish a likelihood of confusion. See, e.g., Network
Automation, Inc. v. Advanced Sys. Concepts, Inc., 638 F.3d 1154 (9th Cir. 2011) (on similar
facts to Rescuecom, reversing district court’s preliminary injunction against Network
Automation, reasoning that Advanced System’s “showing of a likelihood of confusion was
insufficient). Indeed, the Ninth Circuit expressed no opinion on the likelihood of confusion
in Rescuecom. 562 F.3d at 130 (“We have no idea whether Rescuecom can prove that
Google’s use of Rescuecom’s trademark in its AdWords program causes likelihood of
confusion or mistake.”).
202. Rescuecom Corp., 562 F.3d at 125–27.
203. Id. at 127.
204. See Planned Parenthood Fed’n of Am., Inc. v. Bucci, No. 97 Civ 0629 (KMW), 1997
WL 133313, at *3 (S.D.N.Y. Mar. 24, 1997) (“Internet users constitute a national, even
international, audience, who must use interstate telephone lines to access defendant’s web
site on the internet. The nature of the internet indicates that establishing a typical home
page on the internet, for access to all users, would satisfy the Lanham Act’s ‘in commerce’
requirement.”), aff’d, 152 F.3d 920 (2d Cir. 1998); MCCARTHY, supra note 74, at § 25:54.50
(“When an alleged infringing mark is used on the internet, the use is clearly a ‘use in
commerce’ that can be regulated by the Lanham Act under the Constitution.”).
205. See Planned Parenthood Fed’n of Am., Inc., 1997 WL 133313, at *3.
2021] Is This Video Real? 39

directly constitute interstate commerce: “even assuming,

arguendo, that defendant’s activities are not in interstate
commerce for Lanham Act purposes, the effect of those activities
on plaintiff’s interstate commerce activities would place defendant
within the reach of the Lanham Act.” Deepfakes employing the
personas of others, then, can fall within Section 43’s “use in
commerce” requirement, either by way of their mass viewership on
the internet or their effects on the plaintiff.

2. In Connection with Any Goods or Services

Courts have generally read the “in connection with any goods
or services” requirement broadly, applying Section 43(a) to
“defendants furnishing a wide variety of non-commercial public
and civic benefits.” 206 However, courts sometimes apply a
narrower version of it, referring to the “in connection with”
requirement as a “commercial use” requirement, 207 which might be
challenging for a plaintiff to meet in the context of deepfakes. This
narrower interpretation is contradicted by the text of the Lanham
Act, as Section 43(a)(1)(A) (false association) 208 does not limit itself
to infringing uses that are commercial, in contrast to Section
43(a)(1)(B) (false advertising) 209 and Section 43(c) (dilution). 210
Nonetheless, despite Section 43(a)(1)(A)’s silence on the
commerciality of the defendant’s activities, some courts have
imposed this narrower “commercial use” requirement on false
association claims. 211 The Fourth Circuit, for example, has applied
a non-exclusive three factor “commercial speech” test for deciding
whether a defendant’s use of a mark is commercial, which
considers: “whether the speech is an advertisement; whether the

206. United We Stand Am., Inc. v. United We Stand, Am. N.Y., Inc., 128 F.3d 86, 90 (2d
Cir. 1997) (collecting cases); SMJ Grp., Inc. v. 417 Lafayette Rest. LLC, 439 F. Supp. 2d 281,
287 (S.D.N.Y. 2006).
207. See MCCARTHY, supra note 74, at § 23:11.50 (citing, e.g., Keel v. Axelrod, 148 F.
Supp. 3d 411, 419 (E.D. Pa. 2015)).
208. See 15 U.S.C § 1125(a)(1)(A).
209. See 15 U.S.C § 1125(a)(1)(B) (restricting itself to “commercial advertising or
promotion”). See also McCarthy, MCCARTHY, supra note 74, at § 27:71 (similarly noting
there is no requirement of commerciality in Section 43(a)(1)(A)).
210. See 15 U.S.C § 1125(c)(3)(C) (“any noncommercial use of a mark” is “not actionable
as dilution”).
211. See, e.g., Radiance Found., Inc. v. NAACP, 786 F.3d 316, 323 (4th Cir. 2015)
(“Although this case does not require us to hold that the commercial speech doctrine is in
all respects synonymous with the “in connection with” element, we think that doctrine
provides much the best guidance in applying the Act.”).
40 Columbia Journal of Law & Social Problems [55:1

speech references a particular good or service; and whether the

speaker (the alleged infringer) has a demonstrated economic
motivation for his speech.” 212 Although non-profit uses of a mark
can still be found actionable, 213 the commercial use requirement
tends to generally limit the applicability of Section 43(a). 214 Courts
applying a narrower commerciality requirement have rationalized
it as an added protection against intrusion upon First Amendment
values. 215 However, in addition to the fact that a “commercial use”
requirement does not have a basis in the text of Section 43(a) and
many courts do not apply it, 216 it is arguably unnecessary to impose
a “commercial use” requirement for First Amendment purposes,
since courts already apply the very robust Rogers test (as a defense
to infringement) to protect against intrusion on the First
Amendment. 217
Despite the questionable merits of a robust interpretation of the
“in connection with any goods or services” requirement, the fact
that some courts apply it presents a hurdle to Section 43(a) claims

212. Id. (stating “[t]hese are not exclusive factors, and the presence or absence of any of
them does not necessitate a particular result.”) (citing Bolger v. Youngs Drug Products
Corp., 463 U.S. 60, 66–67 (1983)).
213. See Valley Forge Military Acad. Found. v. Valley Forge Old Guard, Inc., 24 F. Supp.
3d 451, 455 (E.D. Pa. 2014) (applying the same three-factor commercial speech test and
granting standing to sue for infringement against a non-profit provider of alumni services).
214. Compare People for Ethical Treatment of Animals v. Doughney, 263 F.3d 359, 365
(4th Cir. 2001) (not applying a commercial use test and allowing infringement action to
proceed against defendant using plaintiff’s mark in its peta.org domain name), with Bosley
Med. Inst., Inc. v. Kremer, 403 F.3d 672, 679 (9th Cir. 2005) (on very similar facts to
Doughney, applying a commercial use test and disallowing an infringement claim based on
defendant’s use of plaintiff’s BOSLEY MEDICAL mark in its BosleyMedical.com domain
name). Of note, the Fourth Circuit, since Doughney, has applied the commercial use test.
See Radiance Found., 786 F.3d at 323. This footnote compares Doughney with Bosley not
to highlight a circuit split, but to show how two very similar cases were decided differently
when the courts applied differing “in connection with” doctrines.
215. See Radiance Found., 786 F.3d at 322 (“At the very least, reading the ‘in connection
with’ element to take in broad swaths of noncommercial speech would be an ‘overextension’
of the Lanham Act’s reach that would ‘intrude on First Amendment values.’”).
216. See, e.g., United We Stand Am., Inc. v. United We Stand, Am. N.Y., Inc., 128 F.3d
86, 90 (2d Cir. 1997); SMJ Grp., Inc. v. 417 Lafayette Rest. LLC, 439 F. Supp. 2d 281, 287
(S.D.N.Y. 2006) (Where the alleged infringing use was in leaflets raising awareness of
restaurant employees’ work conditions, the court held the defendant’s use was in connection
with a public “service,” stating that “defendants’ lack of profit motivation does not place
their activities beyond the scope of the Lanham Act’s definition of ‘services.’”); Brach Van
Houten Holding, Inc. v. Save Brach’s Coal. for Chicago, 856 F. Supp. 472, 475–76 (N.D. Ill.
1994) (collecting cases, and holding that defendant’s “soliciting donations, preparing press
releases, . . . [and engaging in other] work on behalf of its members’ interests constitutes a
‘service’ within the meaning of the Lanham Act.”).
217. See Rogers v. Grimaldi, 875 F.2d 994, 999 (2d Cir. 1989). See also Lemley, supra
note 96, at 1173 (discussing how Rogers limits Lanham Act claims and collecting cases).
2021] Is This Video Real? 41

involving deepfakes. In Radiance Foundation, Inc. v. N.A.A.C.P.,

for example, at issue was whether Radiance Foundation’s article,
entitled, “NAACP: National Association for the Abortion of Colored
People,” infringed the NAACP mark. 218 The article criticized the
NAACP for sponsoring an event Radiance alleged “defied
Christian values.” 219 On the side of the webpage featuring the
article was a “Donate” button allowing viewers to contribute to the
Radiance Foundation, a Christian organization. 220 Despite the
“Donate” button, the Fourth Circuit, applied a narrower,
commercial use-focused version of the “in connection with”
requirement, 221 and held Radiance Foundation’s use of the
NAACP’s marks could not support Lanham Act liability, because
the use was “noncommercial.” 222 The court reasoned that the use
of the NAACP mark “was too attenuated from the donation
solicitation” and did not seem to “denote the recipient of the
donation.” 223
Somewhat similarly, in Bosley Med. Inst., Inc. v. Kremer, the
Ninth Circuit held that “noncommercial use of a trademark as the
domain name of a website” was not actionable under the Lanham
Act. 224 At issue in Bosley was the defendant Kremer’s use of the
plaintiff’s mark in his website’s domain name,
www.BosleyMedical.com, which heavily criticized the plaintiff’s
business. 225 The court characterized defendant’s use of the mark
as noncommercial—”merely . . . to identify the object of consumer
criticism”—and held it was nonactionable. 226 In the vein of cases
like Radiance Foundation and Bosley, a deepfake creator can argue
their use of a plaintiff’s likeness is too removed from any
commercial hook such as advertisements featured on the webpage
or links to other sites offering goods or services.
One common thread of Radiance Foundation and Bosley,
however, that distinguishes them from deepfakes is that confusion

218. Radiance Found., Inc., 786 F.3d at 319.

219. Id. at 320.
220. Id. at 326.
221. Id. at 323 (applying a three-part commercial speech test from Bolger) (citing Bolger
v. Youngs Drug Products Corp., 463 U.S. 60, 66–67 (1983)).
222. Id. at 326.
223. Id. at 327. The court reasoned that “[s]uch a nexus [between the use of the mark
and the good/service] may be present, for example, where the protected mark seems to
denote the recipient of the donation,” which was not true in the case at hand. Id.
224. Bosley Med. Inst., Inc. v. Kremer, 403 F.3d 672, 674 (9th Cir. 2005).
225. Id.
226. Id. at 679.
42 Columbia Journal of Law & Social Problems [55:1

was very unlikely. In Radiance Foundation, the court actually

found there was no likelihood of confusion: “The critical message
conveyed by the satirical mark itself and in the commentary that
follows ensures that no confusion about the source of the
commentary will last, if in fact it is generated at all.” 227 Similarly,
in Bosley, the district court held there was no likelihood of
confusion where the defendant’s webpage, which blatantly
criticized plaintiff Bosley, was unlikely to mislead consumers. 228
Even if a court were to apply the commercial use-focused
interpretation of the “in connection with any goods or services”
requirement discussed above, plaintiffs in deepfake suits have
arguments for bringing deepfakes within the Lanham Act’s
province.
As a threshold matter, deepfakes themselves may constitute
“goods or services” since they are videos from which their creators
can derive profits. 229 Indeed, courts have held that Section 43(a)
“does not require that the mark be used in connection with
separate goods to promote the purchase of those separate goods—
but may be used directly on the defendant’s goods.” 230
In Fifty-Six Hope Road Music, Ltd. v. A.V.E.L.A., Inc., for
example, the Ninth Circuit upheld a jury verdict of false
association, where the defendant used Bob Marley’s image directly
on its merchandise without authorization. 231 In the case of
deepfakes, YouTube, Facebook, and other sites to which deepfakes
may be posted allow video-posters to monetize their content by

227. Radiance Found., Inc., 786 F.3d at 328.

228. Bosley Med. Inst., Inc. v. Kremer, No. Civ.01-1752WQHJMA, 2004 WL 964163, at
*8 (S.D. Cal. Apr. 30, 2004), aff’d in part, rev’d in part and remanded, 403 F.3d 672 (9th Cir.
2005). On appeal, the Ninth Circuit did not reach the issue of likelihood of confusion (or
initial interest confusion, for which there was a good argument), because it found the
defendant’s use was non-infringing as a threshold matter because it was noncommercial.
Bosley Med. Inst., Inc., 430 F.3d at 680 n.3.
229. See, e.g., Naked Cowboy v. CBS, 844 F. Supp. 2d 510, 513 (S.D.N.Y. 2012) (allowing
false endorsement claim to proceed against CBS, where subject of dispute was a YouTube
video) (quoting Fifty-Six Hope Road Music, Ltd. v. A.V.E.L.A., Inc., 778 F.3d 1059, 1072 (9th
Cir. 2015), cert. denied, 577 U.S. 957 (2015)).
230. MCCARTHY, supra note 74, at § 28:15 (“False endorsement cases under Lanham Act
§ 43(a) are not limited to a use of the plaintiff’s identity in advertising to sell goods or
services, but also includes unpermitted use of plaintiff’s identity on a product itself, as with
a name or image on a t-shirt or poster.”).
231. Fifty-Six Hope Road Music, Ltd. v. A.V.E.L.A., Inc., 778 F.3d 1059, 1072 (9th Cir.
2015), cert. denied, 577 U.S. 957 (2015). See also, e.g., Bruce Lee Enterprises, LLC v.
A.V.E.L.A., Inc., No. 10 CV 2333 KMW, 2013 WL 822173 (S.D.N.Y. Mar. 6, 2013) (denying
defendant’s motion for summary judgment on false endorsement based on defendant’s
unauthorized licensing of martial artist Bruce “Lee’s image for use on a variety of products,
including t-shirts, posters, and bobblehead dolls.”).
2021] Is This Video Real? 43

sharing in the revenue generated from advertisements displayed

in conjunction with their videos. 232 Facebook also allows the
owners of Facebook pages to charge for online events and to
generate subscription revenue by providing exclusive content. 233
Unlike in Radiance, the owner of a social media page featuring
deepfakes may, as a direct result of the confusion caused by the
videos, profit via advertising revenue from viewership. Deepfakes
displayed in conjunction with profit-making channels like ad
revenue-sharing are more likely to trigger Section 43(a)’s “in
connection with any goods or services” requirement even under a
narrower formulation of it, because they are economically
motivated. 234 Moreover, some courts have held “[e]ven ‘extremely
minimal’ commercial activity, such as providing a link to a third-
party website that sells goods, can trigger liability.” 235 Thus, the
presentation of deepfakes in conjunction with links to commercial
sites may be sufficient to meet the narrow, commercial use-focused
interpretation of the “in connection with any goods or services”
requirement.
In other contexts, it may be more difficult to establish the “in
connection with any goods or services” element, since deepfakes
can be uploaded by anonymous trolls unassociated with a revenue-
generating social media account or with a website containing
commercial links. However, even deepfakes without a

232. YouTube Partner Earnings Overview, YOUTUBE HELP,

https://support.google.com/youtube/answer/72902?hl=en#zippy=%2Chow-do-i-get-paid
[https://perma.cc/AN3K-MBAP]; How Can I Make Money on Facebook?, FACEBOOK
BLUEPRINT, https://www.facebook.com/business/learn/lessons/how-make-money-facebook
[https://perma.cc/P7A4-354H] [hereinafter FACEBOOK].
233. FACEBOOK, supra note 232.
234. Radiance Found., 786 F.3d at 323 (considering “economic motivation” as one factor
weighing in favor of a finding of “commercial speech” for the purposes of the “in connection
with” goods or services requirement).
235. XPO CNW, Inc. v. R+L Carriers, Inc., No. 16-10391, 2016 WL 4801283, at *3 (E.D.
Mich. Sept. 14, 2016) (quoting Taubman Co. v. Webfeats, 319 F.3d 770, 775 (6th Cir. 2003));
Int’l Bhd. of Teamsters Loc. 651 v. Philbeck, 423 F. Supp. 3d 364, 373–74 (E.D. Ky. 2019)
(denying motion to dismiss where the defendant used the plaintiff’s marks on a Facebook
page in conjunction with links to third-party sites selling goods and services, stating
“Philbeck’s provision of links for discounts to Union members serves to meet the extremely
minimal bar for commercial activity.”); People for Ethical Treatment of Animals v.
Doughney, 263 F.3d 359, 366 (4th Cir. 2001) (finding it favorable to the plaintiff, for the
purposes of establishing the “in connection with” element, that the defendant’s site provided
links “to more than 30 commercial operations offering goods and services,” even though
these links did not direct the viewer to webpages operated by the defendant). See also Bell
v. Worthington City Sch. Dist., No. 2:18-CV-961, 2020 WL 2905803, at *17 (S.D. Ohio June
2, 2020) (“including advertising links on an otherwise non-commercial website falls within
the Lanham Act.”).
44 Columbia Journal of Law & Social Problems [55:1

“commercial” hook as obvious as YouTube advertisement revenue

may still be actionable under other theories of the “in connection
with” provision. Some courts, for example, have allowed
infringement claims to proceed based on an argument that the
defendant’s use of a mark diverted consumers from the plaintiffs’
goods or services. In People for Ethical Treatment of Animals v.
Doughney, at issue was defendant Doughney’s use of the peta.org
domain name for a website he titled, “People Eating Tasting
Animals,” which was a “resource for those who enjoy eating meat,
wearing fur and leather, [and] hunting.” 236 The Fourth Circuit
held:

To use PETA’s Mark “in connection with” goods or services,

Doughney need not have actually sold or advertised goods or
services on the www.peta.org website. Rather, Doughney
need only have prevented users from obtaining or using
PETA’s goods or services, or need only have connected the
website to other’s goods or services. 237

Similarly, deepfakes targeting celebrities may divert consumers

from the celebrities’ profit-making social media profiles or other
activities by damaging their reputations, and deepfakes targeting
political figures may divert funds from their campaigns as a result
of presenting them in a false or misleading manner.
Although a theory of harm based on distortion of the political
process does not fit neatly within the mold of a typical Lanham Act

236. Doughney, 263 F.3d at 363.

237. Id. at 365.; contra Bosley Med. Inst., Inc. v. Kremer, 403 F.3d 672, 679 (9th Cir.
2005). In Bosley, the court stated:
To the extent that the PETA court held that the Lanham Act’s commercial use
requirement is satisfied because the defendant’s use of the plaintiff’s mark as the
domain name may deter customers from reaching the plaintiff’s site itself, we
respectfully disagree with that rationale. . . . We hold today that the
noncommercial use of a trademark as the domain name of a website—the subject
of which is consumer commentary about the products and services represented by
the mark—does not constitute infringement under the Lanham Act.
Id. The Fourth Circuit later distinguished Doughney by stating it “applies specifically
where the infringer has used the trademark holder’s mark in a domain name.” Radiance
Found., Inc., 786 F.3d at 326. However, in Radiance Foundation, as discussed previously,
the court found no likelihood of confusion, so its finding of non-commerciality was not
necessary for its holding. See id. at 328. Moreover, the Fourth Circuit’s limiting of
Doughney to domain names would not prevent a later extension of Doughney to the context
of deepfakes.
2021] Is This Video Real? 45

claim, 238 courts have at times articulated a broad interpretation of

the “in connection with” requirement for the purposes of protecting
the public from confusion during the political process. At issue in
United We Stand America, Inc. v. United We Stand, America New
York, Inc. was a service mark, “United We Stand America,” which
Ross Perot first used in his 1992 presidential campaign, but which
defendant Alex Rodriguez subsequently used in his own political
activities. 239 The Second Circuit held that the defendant’s political
activities constituted “services” for the purposes of the Lanham
Act, reasoning:

Any group trading in political ideas would be free to

distribute publicity statements, endorsements, and position
papers in the name of the “Republican Party,” the
“Democratic Party,” or any other. The resulting confusion
would be catastrophic; voters would have no way of
understanding the significance of an endorsement or position
taken by parties of recognized major names. The suggestion
that the performance of such functions is not within the scope
of “services in commerce” seem to us to be not only wrong but
extraordinarily impractical for the functioning of our political
system. 240

Even though a deepfake’s appropriation of a mark may not directly

benefit the creator of the deepfake, it can lead to similar confusion
among consumers and voters, and, as discussed above, may divert
funding from the target’s political campaign. Moreover, since
there is no requirement in Section 43(a)(1)(A) that the defendant
actually profited for the purposes of establishing infringement, a
claim based on the theory of diversion is not foreclosed by the
statute. 241
In sum, although the “in connection with any goods or services”
requirement of Section 43(a)(1)(A) is perhaps the greatest hurdle
to a claim against the creators of deepfakes, there are ways to link

238. MCCARTHY, supra note 74, at § 5:4 (discussing the protection of consumers and
businesses but not of the political process as the intent of the Lanham Act).
239. See United We Stand Am., Inc. v. United We Stand, Am. N.Y., Inc., 128 F.3d 86, 88
(2d Cir. 1997).
240. Id. at 90.
241. See 15 U.S.C. § 1125. A plaintiff failing to show defendant’s profits would instead
be required to establish her damages resulting from the infringement. See 15 U.S.C.
§ 1117(a).
46 Columbia Journal of Law & Social Problems [55:1

deepfakes to goods and services by establishing the defendant’s

economic motivation or by demonstrating how the deepfake
diverted consumers from the plaintiff’s goods or services.

3. False or Misleading Representation of Fact

The remaining two elements—that of a false or misleading

representation of fact and that of a likelihood of causing consumer
confusion as to the origin, sponsorship, or approval of goods or
services—are easier for deepfakes to meet than the “in connection
with any goods or services” element.
The “misleading representation” element, in particular, is easy
to meet in the context of deepfakes, with courts often glossing over
it. 242 In White v. Samsung Elecs. Am., Inc., for example, Vanna
White, hostess of “Wheel of Fortune” sued Samsung under Section
43(a)(1)(A) for using a “Vanna White robot” in its advertisements
without obtaining her consent. 243 Even though the allegedly
infringing content was a robot depiction of White, and therefore
arguably less of a misrepresentation than if it were a photo (or
deepfake) of White, the Ninth Circuit did not discuss the issue of
misrepresentation and instead focused on the likelihood of
confusion element. 244 In the case of a deepfake, it would be easy to
establish the “misleading representation” element, because
deepfakes are misleading by their very nature. Context or
disclaimers may signal to the viewer the misleading
representation in order to mitigate the likelihood of confusion, but
this is addressed by the likelihood of confusion element rather than
the misleading representation element. 245

242. See, e.g., Bruce Lee Enterprises, LLC v. A.V.E.L.A., Inc., No. 10 CV 2333 KMW,
2013 WL 822173, at *3, *19 (S.D.N.Y. Mar. 6, 2013) (denying defendant’s motion for
summary judgment on false endorsement, and glossing over the “false or misleading
representation” element, where defendant engaged in unauthorized licensing of martial
artist Bruce “Lee’s image for use on a variety of products, including t-shirts, posters, and
bobblehead dolls.”).
243. White v. Samsung Elecs. Am., Inc., 971 F.2d 1395, 1396 (9th Cir. 1992) (“The ad
depicted a robot, dressed in a wig, gown, and jewelry which Deutsch consciously selected to
resemble White’s hair and dress. The robot was posed next to a game board which is
instantly recognizable as the Wheel of Fortune game show set, in a stance for which White
is famous.”). See also Wendt v. Host Int’l, Inc., 125 F.3d 806 (9th Cir. 1997) (reversing
summary judgment against plaintiff actors on false endorsement where defendant sold
robots based on actors’ likenesses).
244. See id. at 1399.
245. MCCARTHY, supra note 74, at § 19:72.
2021] Is This Video Real? 47

4. Likelihood of Confusion

False association’s likelihood of confusion element originates

from the portion of Section 43(a) prohibiting false or misleading
representations “likely to cause confusion, or to cause mistake, or
to deceive as to the affiliation, connection, or association of such
person with another person.” 246 As discussed infra in Part II.A,
confusion is the “keystone” and “heart” of a claim under the
Lanham Act. 247
Courts generally apply multi-factor tests, which differ slightly
based on the circuit, to find a likelihood of confusion. 248 In
Downing v. Abercrombie & Fitch, the Ninth Circuit crafted a test
specific to false association claims, which considers the following
eight factors:

1) The level of recognition that the plaintiff has among the

segment of the society for whom the defendant’s product
is intended;
2) The relatedness of the fame or success of the plaintiff to
the defendant’s product;
3) The similarity of the likeness used by the defendant to
the actual plaintiff;
4) Evidence of actual confusion;
5) Marketing channels used;
6) Likely degree of purchaser care;
7) Defendant’s intent on selecting the plaintiff; and
8) Likelihood of expansion of the product lines. 249

246. 15 U.S.C. § 1125(a).

247. MCCARTHY, supra note 74, at § 23:1; Allen v. Nat’l Video, Inc., 610 F. Supp. 612,
627 (S.D.N.Y. 1985) (quoting Standard & Poor’s Corp. v. Commodity Exch., Inc., 683 F.2d
704, 708 (2d Cir. 1982)).
248. See, e.g., Polaroid Corp. v. Polarad Elecs. Corp., 287 F.2d 492, 128 (2d Cir. 1961)
(providing the basis for the Second Circuit Polaroid likelihood of confusion factors). See also
MCCARTHY, supra note 74, at § 23:19 (discussing the different factors courts use to find a
likelihood of confusion).
249. Downing v. Abercrombie & Fitch, 265 F.3d 994, 1007–08 (9th Cir. 2001) (citing
AMF Inc. v. Sleekcraft Boats, 599 F.2d 341, 348–49 (9th Cir. 1979) (retooling the Sleekcraft
factors, which the Ninth Circuit traditionally applies to find a likelihood of confusion, for
the context of a false endorsement claim). Modified versions of the Downing test have also
been adopted by other circuits. See Facenda v. N.F.L. Films, Inc., 542 F.3d 1007, 1020 (3d
Cir. 2008); MCCARTHY, supra note 74, at § 28:15.
48 Columbia Journal of Law & Social Problems [55:1

These factors are highly case-specific and are neither of equal

importance nor applicable to every case. 250 So, this Part will focus
on those factors most likely to raise difficulties when applied to
deepfakes. The first factor, “level of recognition,” in particular, has
the potential to limit false association claims to those by public
figures, 251 but courts have not strictly required “celebrity” status
to bring false association actions: “there is a level of consumer
recognition short of celebrity—as the term is usually understood—
capable of causing consumer confusion.” 252 Despite this, private
citizens generally unknown to the public may have trouble
establishing infringement claims for at least two reasons: viewers
are less likely to be deceived into watching a deepfaked video
featuring unknown individuals, and viewers are unlikely to be
confused by a persona that is unrecognizable and therefore not
meaningful to them. 253
Additionally, the second factor, “relatedness of the fame or
success of the plaintiff to the defendant’s product,” is not obviously
applicable to deepfakes, where the “product” is the fake video itself.
Regardless, courts have found a likelihood of confusion in false
association cases when there was no obvious connection between
the plaintiff’s fame and the defendant: in Waits v. Frito-Lay, for
example, where defendant’s Doritos commercial featured an
allegedly infringing voice-impersonation of Tom Waits, the court
affirmed a jury finding of a likelihood of confusion even though
there was no connection between Tom Waits’ fame and flavored
tortilla chips. 254 When courts actually do analyze the relatedness
factor, they seem to interpret it broadly, perhaps because public

250. Fifty-Six Hope Road Music, Ltd. v. A.V.E.L.A., Inc., 778 F.3d 1059, 1069 (9th Cir.
2015) (applying the Downing factors to find a likelihood of confusion as to whether Bob
Marley endorsed a sportswear company that used his likeness on its T-shirts).
251. MCCARTHY, supra note 74, at § 28:17 (“Applying the first factor in the Ninth
Circuit’s Downing test, a noncelebrity’s image would probably have a very low level of
recognition among the target audience of millions of persons in the consumer market for
defendant’s goods or services.”).
252. Bondar v. LASplash Cosmetics, No. 12 Civ. 1417(SAS), 2012 WL 6150859, *7
(S.D.N.Y. Dec. 11, 2012) (denying defendant’s motion to dismiss plaintiff model’s false
endorsement claim, even though it was “unlikely the public at large is familiar” with the
plaintiff).
253. See Electra v. 59 Murray Enterprises, Inc., No. 19-235, 2021 WL 438900, at *18 (2d
Cir. Feb. 9, 2021) (reasoning that that the plaintiff—an unknown model—did not have a
sufficiently recognized persona to establish a likelihood of confusion). See also MCCARTHY,
supra note 74, at § 28:17 (discussing how the likelihood of confusion test is likely to prevent
many non-celebrities from bringing Section 43(a) false endorsement claims).
254. Waits v. Frito-Lay, Inc. 978 F.2d 1093, 1111 (9th Cir. 1992) (affirming jury finding
of likelihood of confusion and verdict of false endorsement).
2021] Is This Video Real? 49

figures have the ability to commercially exploit their fame in a

variety of different manners. In White v. Samsung Elecs. Am., Inc.,
for example, the court found White’s fame as a gameshow host was
related to the defendant’s VCR product, because VCRs can be used
to record gameshows. 255 A target of deepfake, such as a political
figure, can similarly argue their fame, derived from their public
statements and political activities, is related to the defendant’s
product—a distorted video of such statements and activities.
In addition to the above factors, courts would likely consider
whether the false representation in the deepfake is obvious from
the context 256 or if the defendant included a disclaimer, 257 the lack
of which would potentially weigh against defendants under the
fourth factor—”actual confusion”—and the seventh factor—
”defendant’s intent.” In sum, the likelihood of confusion analysis
is highly fact-dependent, but deepfakes generally have a high
propensity to cause confusion due to how realistic they can appear.
While this Part has outlined the elements of a prima facie case
against deepfake creators—use in commerce, in connection with
goods or services, false or misleading representation of fact, and
likelihood of confusion—noting the difficulties in particular of
establishing the “in connection with any goods or services”
element, litigants must also be aware of potential First
Amendment defenses and the ability to secure judgments against
third parties.

C. FIRST AMENDMENT DEFENSE

Deepfake creators are likely to raise a First Amendment

defense, arguing their videos are expressive works or parodies
entitled to First Amendment protections. 258 This Note argues that
if a deepfake does not include disclaimers and is effective at
deceiving viewers, the creator is unlikely to have a strong First

255. White v. Samsung Elecs. Am., Inc., 971 F.2d 1395, 1400 (9th Cir. 1992).
256. See Roberts v. Bliss, 229 F. Supp. 3d 240, 252 (S.D.N.Y. 2017) (granting defendant’s
motion to dismiss false endorsement claims and reasoning that defendant’s ad, which
“superimpose[d] graphic images of life-sized appetizers over Plaintiff’s entire image,” was
so “outlandish” that a reasonable viewer would recognize it as a parody of plaintiff’s video).
257. See Allen v. Nat’l Video, Inc., 610 F. Supp. 612, 628 (S.D.N.Y. 1985) (“The failure of
defendant National to include any disclaimer on all but one of the uses of the photograph
also supports a finding of, at best, dubious motives.”).
258. See MCCARTHY, supra note 74, at § 28:16 (“The Rogers test is a balancing of private
and public interests when a mark is used in an expressive work. It is not applicable to
“commercial” works such as a traditional advertisement or an infomercial.”).
50 Columbia Journal of Law & Social Problems [55:1

Amendment defense. To determine whether an expressive use of

a mark (e.g., a persona) is protected, courts apply the two-part
Rogers test, which protects a defendant’s use unless the alleged use
(i) has “no artistic relevance” to the disputed work or, (ii) if there
is artistic relevance, the use “explicitly misleads as to the source or
the content of the work.” 259 Courts interpret the second element,
“explicitly misleads,” as a stricter version of the likelihood of
confusion test, requiring the argument for confusion to be
“particularly compelling.” 260
In Warner Brothers Entertainment v. Global Asylum, Inc., for
example, the Ninth Circuit affirmed a finding that the defendant’s
“Age of Hobbits” film, which was about early humans that
scientists nicknamed “Hobbits,” was explicitly misleading and
infringing of plaintiff’s “Hobbit” trademark. 261 Particularly
disfavorable to the defendant was the court’s finding that the
defendant took specific efforts to deceive viewers and associate the
film with the plaintiff by (i) failing to advertise or otherwise inform
potential viewers that the film was about early humans rather
than Hobbits from J.R.R. Tolkien’s folklore, (ii) releasing the film
three days before Warner Brothers released the “real” Hobbit
movie, and (iii) failing to correct media coverage that described the
film as “a reimagined version of J.R.R. Tolkien’s mythical
universe.” 262 Even assuming a deepfake’s use of a plaintiff’s
persona is artistically relevant to the content of the deepfake, most
deepfakes are likely to “explicitly” mislead in a manner analogous
to that of the “Age of Hobbits” film. Like the defendants in Global
Asylum, deepfake creators go out of their way to craft highly
realistic videos that are likely to deceive their viewers.
Deepfake creators may argue that even if their videos initially
create confusion, the viewer will eventually deduce the fakery.
Indeed, in Cliffs Notes, Inc. v. Bantam Doubleday Dell Pub. Group,
Inc., where the defendant sold parodic “Spy Notes” books featuring
the “distinctive yellow color, black diagonal stripes and black

259. Rogers v. Grimaldi, 875 F.2d 994, 999 (2d Cir. 1989).
260. Twin Peaks Prods., Inc. v. Publ’ns Intern., Ltd., 996 F.2d 1366, 1379 (2d Cir. 1993).
See also 6 J. THOMAS MCCARTHY, MCCARTHY ON TRADEMARKS AND UNFAIR COMPETITION
§ 31:144.50 (5th ed. 2021) (discussing the Rogers test and collecting cases from different
circuits applying it).
261. Warner Bros. Ent. v. Glob. Asylum, Inc., No. CV 12-9547 PSG (CWx), 2012 WL
6951315, at *17 (C.D. Cal. Dec. 10, 2012), aff’d sub nom. Warner Bros. Ent., Inc. v. Glob.
Asylum, Inc., 544 F. App’x 683 (9th Cir. 2013).
262. Id. at *13.
2021] Is This Video Real? 51

lettering of Cliffs Notes,” the Second Circuit held there was no

likelihood of confusion and therefore the defendant’s use could not
be explicitly misleading under Rogers. 263 The court reasoned that
even if consumers were confused by the cover, they would most
likely open the books and identify them as parodies before
purchasing them. 264 In the same vein, deepfake videos may
initially mislead their viewers but, upon being viewed in their
entirety, may be obviously fake.
Many deepfakes, however, are distinguishable from the books
discussed in Cliffs Notes: a deepfake viewer—i.e., an internet
user—may not watch enough of a deepfake to identify it as fake,
and, more significantly, unlike the Spy Notes books, a deepfake
may have zero disclaimers, whether explicit or contextual, that it
is fake. Unless the fakery is deducible from the context of the video
or there is a disclaimer, 265 the creators of pernicious deepfakes are
unlikely to have a strong First Amendment defense under the
Rogers test.

D. SECONDARY LIABILITY

As discussed infra in Part I, prominent websites like Facebook

and YouTube have policies prohibiting pernicious deepfakes and
have pursued methods of detecting deepfakes uploaded to their
sites. 266 The owners of less resourced or less scrutinized “fringe”
websites, however, may choose to ignore deepfakes that are posted
on their platforms. As many deepfakes are uploaded to the
internet by anonymous trolls who are difficult to trace for the
purposes of a lawsuit, it may be critical for deepfake targets to be
able to sue website owners. 267 Unlike in the case of other causes of
action like defamation, discussed supra in Part II, false association

263. Cliffs Notes, Inc. v. Bantam Doubleday Dell Publ’n Grp., Inc., 886 F.2d 490, 497 (2d
Cir. 1989).
264. Id. at 496.
265. See MCCARTHY, supra note 74, at § 31:154 (Judges will generally grant First
Amendment defenses and reject infringement claims where “the joke is readily apparent
because of the fame of the target mark and/or differences in the overall setting of the
parody.”).
266. See Bickert, supra note 64; Fake Engagement Policy, YOUTUBE HELP,
https://support.google.com/youtube/answer/3399767?hl=en [https://perma.cc/5W7M-
GG6H].
267. Before pursuing a claim against a website owner, a plaintiff may seek to subpoena
a third-party such as a domain registrar to secure the identity of an anonymous defendant
who, for example, owns a website that is posting infringing material. See, e.g., Marketo,
Inc. v. Doe, No. 18-CV-06792-JSC, 2018 WL 6046464, at *1 (N.D. Cal. Nov. 19, 2018).
52 Columbia Journal of Law & Social Problems [55:1

claims against website owners on theories of secondary liability are

not foreclosed by Section 230 of the Communications Decency
Act. 268 While plaintiffs may additionally pursue a direct claim of
false association against website owners, this section discusses the
hurdles of pursuing a claim of contributory false association.

1. Section 230 of the Communications Decency Act

To establish a contributory false association claim against a

website hosting confusing deepfakes uploaded by a third-party
user, a plaintiff may need to respond to a challenge that Section
230 of the Communications Decency Act bars such a claim. As will
be discussed below, both precedent and historical characterization
of the law of false association indicate the claim is not barred.
Section 230(c) of the CDA reads: “No provider or user of an
interactive computer service shall be treated as the publisher or
speaker of any information provided by another information
content provider.” 269 This effectively gives websites like Facebook,
YouTube, and other social media platforms immunity from causes
of action based their publication of material originating from a
third party. 270 Such immunity is the reason why Facebook cannot
be liable for defamation when a third party user posts a
defamatory message on its platform. 271
The general rule of immunity, however, is accompanied by
exceptions in Section 230(e), which stipulates that the statute does
not foreclose actions based on, among other things, intellectual
property law. 272 Indeed, Section 230(e)(2) states, “Nothing in this
section shall be construed to limit or expand any law pertaining to
intellectual property,” 273 i.e., the CDA does not immunize websites
from secondary liability for intellectual property claims. 274 Courts
have interpreted Section 230(e)(2) to allow Lanham Act claims to
proceed against website owners, with only a couple of exceptions:
some courts—the Ninth and D.C. Circuits—have held that false

268. See 47 U.S.C. § 230; MCCARTHY, supra note 74, at § 25A:42.50 (collecting cases)
(“Since a trademark is clearly a type of “intellectual property,” CDA statutory immunity
does not shield such claims. All courts agree that a claim under federal law for trademark
infringement is an intellectual property claim not precluded by the CDA.”).
269. 47 U.S.C. § 230(c).
270. See 47 U.S.C. § 230; MCCARTHY, supra note 74, at § 25A:42.50.
271. See id; MCCARTHY, supra note 74, at § 25A:42.50.
272. 47 U.S.C. § 230(e)(2).
273. Id.
274. MCCARTHY, supra note 74, at § 25A:42.50.
2021] Is This Video Real? 53

advertising is not an “intellectual property” claim under Section

230(e)(2) and, therefore, that Section 230 bars contributory false
advertising claims against website owners. 275 Similarly, the
District Court for the Western District of Washington has held that
false designation of origin is not an intellectual property claim and
that such a claim is also barred by Section 230. 276 These holdings
are non-obvious, however, and the District Court for the Southern
District of New York has taken the opposite view, holding that
Section 230 bars neither false advertising nor false designation of
origin claims. 277
More importantly, and relevant to this Note, no court has held
that Section 230 bars false association claims. 278 Indeed the courts
to have addressed the issue, whether in dicta 279 or in their
holding, 280 have taken the position that false association claims are
not barred by Section 230. Precedent aside, treating the false
association claim as arising under intellectual property law for the
purposes of Section 230 is consistent with how courts and
commentators have historically characterized the claim: “Courts
hold that in the context of § 43(a)(1)(A), a human persona or
identity is a kind of ‘trademark’ which is infringed by a false
endorsement [i.e., false association].” 281 This is different than false
advertising and false designation of origin—the Lanham Act
claims that some courts have held do not fall in the intellectual
property law exception to Section 230 immunity. Unlike false

275. Enigma Software Grp. USA, LLC v. Malwarebytes, Inc., 946 F.3d 1040, 1053 (9th
Cir. 2019); Marshall’s Locksmith Serv. Inc. v. Google, LLC, 925 F.3d 1263, 1269 (D.C. Cir.
2019).
276. Corker v. Costco Wholesale Corp., No. C19-0290RSL, 2019 WL 5895430, at *6 (W.D.
Wash. Nov. 12, 2019) (barring false designation of origin claim under Section 230).
277. Gucci Am., Inc. v. Hall & Assocs., 135 F. Supp. 2d 409, 412 (S.D.N.Y. 2001).
278. Of note, in Corker v. Costco Wholesale Corp., the court refers to the plaintiff’s claim
as “false association” and as “false association of origin,” in addition to “false designation of
origin,” even though the facts pleaded indicate the claim was based on a theory of false
designation of origin. No. C19-0290RSL, 2019 WL 5895430, at *4 n.4 (W.D. Wash. Nov. 12,
2019).
279. Stayart v. Yahoo! Inc., 651 F. Supp. 2d 873, 885 (E.D. Wis. 2009), aff’d, 623 F.3d
436 (7th Cir. 2010) (stating in dicta, “For example, if Stayart successfully stated a claim for
false endorsement under the Lanham Act, Yahoo! would not be immune from liability for
those claims because such a claim would probably be considered an intellectual property
claim.”); Nieman v. Versuslaw, Inc., No. 12-3104, 2012 WL 3201931, at *8 (C.D. Ill. Aug. 3,
2012), aff’d, 512 F. App’x 635 (7th Cir. 2013) (in dicta, stating Section 230 would not bar
plaintiff’s false association claim, if adequately pled).
280. Doe v. Friendfinder Network, Inc., 540 F. Supp. 2d 288, 306 (D.N.H. 2008) (holding
that the CDA immunized defendant from state law tort claims but allowing false association
claim to proceed, unbarred by the CDA).
281. MCCARTHY, supra note 74, at § 28:15 (citation omitted).
54 Columbia Journal of Law & Social Problems [55:1

advertising and false designation of origin—which are arguably

based solely on false representations themselves rather than on
the appropriation of intellectual property rights—false association
is premised on the plaintiff’s intellectual property right in her
persona. 282 For these reasons, there is strong support for the
position that Section 230 does not bar false association claims
against website owners.

2. Contributory Infringement

Even though the CDA does not grant websites immunity from
Lanham Act claims, plaintiffs still must establish contributory
infringement in order to hold a website owner accountable, in
addition to first establishing direct infringement. 283 Contributory
trademark infringement requires a plaintiff to show the defendant
website owner had actual or constructive knowledge of
infringement and nonetheless continued to supply its services to
the infringer. 284 In Louis Vuitton Malletier, S.A. v. Akanoc, for
example, the Ninth Circuit affirmed a jury verdict of contributory
infringement where the plaintiff showed that the defendant server
operator was aware of infringements on websites it hosted and
failed to remove the infringing content in response to at least
eighteen Notice of Infringement letters from the plaintiff. 285 In
contrast, in Tiffany (NJ) Inc. v. eBay, Inc., the Second Circuit
upheld a finding of no contributory infringement, because Tiffany

282. See Enigma Software Grp. USA, LLC v. Malwarebytes, Inc., 946 F.3d 1040, 1053–
54 (9th Cir. 2019):
Enigma’s Lanham Act claim derives from the statute’s false advertising provision.
Enigma alleges that Malwarebytes mischaracterized Enigma’s most popular
software programs in order to divert Enigma’s customers to Malwarebytes. These
allegations do not relate to or involve trademark rights or any other intellectual
property rights. Thus, Enigma’s false advertising claim is not a claim “pertaining
to intellectual property law” within the meaning of § 230(e)(2). The district court
correctly concluded that the intellectual property exception to immunity does not
encompass Enigma’s Lanham Act claim.
283. Id. at §§ 25:20.50, 25:17 (collecting cases).
284. See Tiffany (NJ) Inc. v. eBay Inc., 600 F.3d 93, 110 (2d Cir. 2010) (rejecting a
contributory trademark infringement claim by Tiffany against eBay where the plaintiff
could not establish whether eBay knew of, and failed to act on, specific infringing uses of
Tiffany marks); Louis Vuitton Malletier, S.A. v. Akanoc Sols., Inc., 658 F.3d 936, 942 (9th
Cir. 2011) (“To prevail on its claim of contributory trademark infringement, Louis Vuitton
had to establish that Appellants continued to supply its services to one who it knew or had
reason to know was engaging in trademark infringement.”).
285. Louis Vuitton Malletier, 658 F.3d at 940.
2021] Is This Video Real? 55

did not show eBay had knowledge of specific infringing uses. 286
eBay’s general knowledge of counterfeit Tiffany goods and the fact
a “significant portion of the ‘Tiffany’ sterling silver jewelry listed
on the eBay website . . . was counterfeit” was not enough to
establish contributory infringement. 287
Although contributory infringement’s strict requirement of
actual or constructive knowledge is a hurdle, at least the CDA does
not immunize websites from contributory false association claims
as it does from secondary liability based on other causes of action
like defamation. 288 Regardless of the causes of action they
ultimately pursue, deepfake targets are most likely to sue only
after first pursuing cheaper and more efficient avenues of
removing their deepfake from online circulation, such as reporting
the videos to the social media sites hosting them. Large,
mainstream websites like Facebook and YouTube, as discussed
above, have indicated their intentions to protect the public from
deepfakes and are therefore more likely to comply with the targets
of pernicious deepfakes. Therefore, deepfake targets are most
likely to pursue contributory trademark infringement against
fringe or less resourced website operators that refuse to comply
with Notice of Infringement letters and willfully ignore pernicious
and infringing deepfakes uploaded to their websites. As
contributory false association claims against website owners are
not barred by the CDA, a false association theory may be one of
very few options available to deepfake targets who have exhausted
non-legal solutions.

CONCLUSION

As the discussion of contributory infringement above indicates,

the courtroom is one of many avenues—including third-party
detection software and cooperation among social media sites—
useful for combatting pernicious deepfakes. This Note has
provided an additional tool to the arsenal of those seeking
remedies to the harms created by deepfakes. Particularly, this
Note has made the case that the law of false association is well-
suited to address deepfakes due to its focus on confusion and,
crucially, the fact contributory infringement claims are not barred

286. Tiffany (NJ) Inc., 600 F.3d at110.

287. Id. at 98.
288. See MCCARTHY, supra note 74, at § 22:4.50.
56 Columbia Journal of Law & Social Problems [55:1

by the Communications Decency Act. Although false association’s

standing and “in connection with any goods or services”
requirements may pose hurdles to claims against creators of
deepfakes unlinked to advertising, this Note has highlighted
theories of both requirements under which a plaintiff could
arguably proceed.
A false association cause of action against certain deepfake
creators may strain existing notions of the Lanham Act. However,
deepfake technologies and social media have evolved such that
individuals can now profit from creating misleading depictions of
others and posting them online—a mischief this Note argues is
contemplated by the law of false association. Just as the Lanham
Act has adapted to the technological advances of the internet and
Photoshop, 289 it may adapt to the rise of deepfakes.

289. See Geiger v. C&G of Groton, Inc., 424 F. Supp. 3d 276, 285 (D. Conn. 2019) (holding
that the plaintiffs successfully stated a Lanham Act false advertising claim where the
defendant altered images of the plaintiffs in order to make it appear that the plaintiffs were
affiliated with the defendants’ strip clubs).