Concept Paper

Generative Artificial Intelligence and Regulations: Can We Plan a

Resilient Journey Toward the Safe Application of Generative
Artificial Intelligence?
Matteo Bodini

Department of Economics, Management and Quantitative Methods, University of Milan,

Via Conservatorio 7, 20122 Milan, Italy; [email protected]

Abstract: The rapid advancements of Generative Artificial Intelligence (GenAI) technologies, such as
the well-known OpenAI ChatGPT and Microsoft Copilot, have sparked significant societal, economic,
and regulatory challenges. Indeed, while the latter technologies promise unprecedented productivity
gains, they also raise several concerns, such as job loss and displacement, deepfakes, and intellectual
property violations. The present article aims to explore the present regulatory landscape of GenAI
across the major global players, highlighting the divergent approaches adopted by the United States,
United Kingdom, China, and the European Union. By drawing parallels with other complex global
issues such as climate change and nuclear proliferation, this paper argues that the available traditional
regulatory frameworks may be insufficient to address the unique challenges posed by GenAI. As a
result, this article introduces a resilience-focused regulatory approach that emphasizes aspects such
as adaptability, swift incident response, and recovery mechanisms to mitigate potential harm. By
analyzing the existing regulations and suggesting potential future directions, the present article aims
to contribute to the ongoing discourse on how to effectively govern GenAI technologies in a rapidly
evolving regulatory landscape.

Keywords: GenAI; GenAI regulations; GenAI regulatory framework; resilience

Citation: Bodini, M. Generative

Artificial Intelligence and Regulations:
1. Introduction
Can We Plan a Resilient Journey
Toward the Safe Application of
The latest advancements of Generative Artificial Intelligence (GenAI) technologies,
Generative Artificial Intelligence?
such as ChatGPT, Microsoft Copilot, Google Gemini, and many others have led to the cre-
Societies 2024, 14, 268. https://
ation of conversations capable of miming human interactions, generating realistic images,
doi.org/10.3390/soc14120268 engineering of advanced computer code, and execution of several other human-related
tasks [1]. However, the latter progress has raised several concerns, such as potential issues
Academic Editor: Theodora Saridou,
related to future job loss and displacement [2], the rise of deepfakes [3], and violations
Charalampos Dimoulas
of intellectual property rights [4]. On the other hand, GenAI technologies could signif-
Received: 3 November 2024 icantly boost productivity, potentially driving growth for businesses and economies [5].
Revised: 9 December 2024 For instance, Bank of America predicted that by the year 2030, GenAI could bring USD
Accepted: 13 December 2024 15 trillion into the global economy [6]. Given the above-mentioned implications, it is no
Published: 18 December 2024 surprise that regulatory agencies are trying to take actions against GenAI. Indeed, as per
the Stanford Artificial Intelligence (AI) Index Report, over 140 laws related to AI were
enacted by the year 2023 across world states (refer to Figure 1) [7]. Despite the introduced
laws and regulations, a major question still remains, i.e., what form of regulations are the
Copyright: © 2024 by the author.
most effective in addressing the recent challenges posed by GenAI technologies, across mul-
Licensee MDPI, Basel, Switzerland.
tiple sectors? The present concept paper adopts a cross-sectoral perspective, recognizing
This article is an open access article
distributed under the terms and
that regulatory challenges and demands may vary significantly across multiple domains,
conditions of the Creative Commons
such as education, healthcare, politics, and the military. Indeed, the resilience-based re-
Attribution (CC BY) license (https://
search design provided herein is aimed to overcome domain-related discrepancies, thereby
creativecommons.org/licenses/by/ delivering a flexible yet robust approach applicable across multiple sectors.
4.0/).

Societies 2024, 14, 268. https://doi.org/10.3390/soc14120268 https://www.mdpi.com/journal/societies

Societies 2024, 14, 268 2 of 16

Figure 1. The figure reports the number of bills passed into law by national legislative bodies
(e.g., congress and parliament) with the keyword “artificial intelligence” (translated to the respec-
tive languages) in the title or body of the bill. The data reported in the figure were collected
from the 2023 Stanford AI Index Report [7] (p. 378). The top five world countries for passed
AI-related bills (whose number of bills is here, respectively, reported for each state within paren-
thesis) are the United States (US) (23), Portugal (15), Belgium (12), Spain (11), and South Korea
(10). The figure was adapted from Roser [8], with minor processing, from the 2023 Stanford AI
Index Report [7] (p. 376), under the terms of the Creative Commons Attribution License—CC BY 4.0
(https://creativecommons.org/licenses/by/4.0/, accessed on 6 December 2024).

Regarding the above-mentioned term of “resilience”, within the context of GenAI

technologies, such a term refers to the capability of systems to withstand, adapt to, and
recover from disruptions while maintaining their functionality. Drawing from the “Ethics
Guidelines for Trustworthy AI” provided by the High-Level Expert Group on AI, resilience
is a crucial component of technically robust and safe systems, encompassing features such
as fault tolerance, security against adversarial attacks, and recovery mechanisms [9]. The
latter concept also aligns with the insights of the European Union Agency for Cybersecurity
(ENISA) into AI cybersecurity, emphasizing the importance of preemptive risk identifi-
cation, incident mitigation, and post-incident recovery [10]. Therefore, a resilient GenAI
system must not only integrate preventive safeguards but also ensure dynamic adaptability
to evolving challenges, safeguarding its reliability across diverse applications.
The complexity of regulating GenAI lies in its multiple available applications and the
range of risks it introduces. Current regulatory efforts have primarily focused on issues
such as data privacy, security, and transparency. For instance, the European Union’s (EU)
proposed AI Act sets forth a risk-based framework that classifies AI systems according to
the potential harm they could cause, ranging from minimal to unacceptable risk [11,12].
High-risk systems, such as those used in critical infrastructure or law enforcement, would
face stricter regulatory scrutiny, including requirements for transparency, oversight, and
accountability. However, are the already available regulations enough to address the chal-
lenges specific to the latest advancements brought by GenAI? It must be acknowledged
that, while the present work predominantly examines novel regulatory initiatives, not all
the challenges introduced by GenAI are entirely new. Indeed, depending on the considered
domain, pre-existing regulatory frameworks, such as those related to the International
Humanitarian Law (IHL), have shown remarkable adaptability in addressing emerging
technological challenges. For instance, the IHL framework, as discussed in Davison [13],
demonstrated how established principles, such as distinction and proportionality, have
been applied successfully to regulate autonomous weapon systems which, similarly to
GenAI, are characterized by rapid technological advancements. Drawing from such an
example, certain aspects of GenAI regulation could benefit from leveraging existing laws
Societies 2024, 14, 268 3 of 16

and principles, thereby reducing the need for entirely new frameworks. However, such a
reflection even underscores the dual necessity of both evaluating the sufficiency of existing
frameworks while considering the unique attributes of GenAI that demand innovative
regulatory responses. Thus, while the present work acknowledges the remarkable insights
offered by established frameworks, it points out that the latter ones alone may not be suffi-
cient to address all the GenAI-related challenges. As a final consequence, the latter point
underscores the necessity for the development of novel and comprehensive frameworks
specifically designed to meet all the unique challenges posed by Generative AI.
In the past years, several researchers attempted to categorize the designed regulatory
AI and GenAI frameworks within different world countries relying on the degree of legally
binding and non-binding regulations, and whether such regulations were sector-specific or
could be applied to the entire considered state economy [14]. For instance, the Executive
Order on AI issued by President J. Biden in the US in October 2023 represents a federal
initiative to establish safety and security standards, potentially guiding the US toward more
rigorous restrictions [12,15]: 14 US states as of 2024 approved some sort of AI regulation
mainly focused on consumer and/or data privacy, e.g., the Illinois AI Video Interview
Act [16]. The government of the United Kingdom defined several principles (safety, security
and robustness, appropriate transparency and explainability, fairness, accountability and
governance, and contestability and redress) and a central function to assess regulation and
monitor AI and GenAI technologies [17,18]. The People’s Republic of China introduced
regulations on recommendation algorithms [19] (in the year 2022), deep synthesis [20]
(in the year 2023), and GenAI [21] (in the year 2023) [22]. The previously mentioned EU
AI Act (released in the year 2024) states requirements based on risks induced by AI and
GenAI systems. In particular, applications denoted with high risks require an assessment
before release [11,12]. When considering the balance of interests, it can be observed that
the US leans toward fostering innovation while the EU prioritizes citizen welfare and
China emphasizes full control over the state [23]. As a final point, it must be finally noted
that the majority of countries of all the world continents (except Africa) developed, or are
developing, national AI strategies (refer to Figure 2). For the sake of readability, Table 1
summarizes the key aspects of the above-described regulations.

Figure 2. The figure reports world countries with national AI strategies (in light blue color), countries
where AI strategies are under development (in grey color), and countries where AI strategies are
not released or not already planned (in orange color). The data reported in the figure were collected
from the 2023 Stanford AI Index Report [7] (p. 392). The figure was adapted from Roser [8],
with minor processing, from the 2023 Stanford AI Index Report [7] (p. 391), under the terms of the
Creative Commons Attribution License—CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/,
accessed on 6 December 2024).
Societies 2024, 14, 268 4 of 16

Table 1. Overview of AI and GenAI regulatory frameworks for the analyzed countries or world
regions, categorized by the degree of stringency and focus. The table outlines key regulations and
taken actions and highlights each country’s or region’s main regulatory focus.

Country/Region Regulatory Approach Key Regulations Main Focus

Executive Order on AI
(2023) [15], state-level
Legally non-binding, Innovation, safety,
US regulations (e.g., Illinois
evolving and security
AI Video Interview
Act) [16]
Principles for AI
regulation: safety,
Legally non-binding, Balancing innovation
United Kingdom transparency, fairness,
with oversight with safety
accountability, and
monitoring [17]
Regulations on
algorithms (2022) [19],
People’s Republic Legally binding, State control
deep synthesis [20]
of China sector-specific and oversight
(2023), and GenAI [21]
(2023)
EU AI Act (2024),
requiring risk
Legally binding, Citizen welfare
EU assessments for
risk-based and protection
high-risk AI
applications [11]

2. Research Methodology
This article relied on a qualitative research methodology to explore the regulatory
landscape of GenAI and, within the forthcoming sections, it introduces a resilience-focused
regulatory framework. In particular, the designed research process involved the following
key steps:
• Literature review: A comprehensive review of the existing literature on GenAI tech-
nologies, their societal and economic impacts, and current regulatory approaches was
conducted. The latter one included academic papers, industry reports, policy docu-
ments, and regulatory guidelines from major global players such as the US, United
Kingdom, China, and the EU.
• Comparative analysis: This study compared the regulatory frameworks of different
countries and regions to identify commonalities and differences in their approaches
to GenAI governance. Such a comparative analysis helped in understanding the
strengths and weaknesses of various regulatory models and informed the development
of the proposed resilience-focused framework.
• Thematic analysis: Key themes related to GenAI regulation, such as data privacy,
security, transparency, and resilience, were identified and analyzed. The thematic
analysis provided insights into the critical aspects that need to be addressed in a robust
regulatory framework.
• Case studies: Specific case studies of regulatory responses to other complex global
challenges, such as climate change, nuclear proliferation, and space debris, were
examined. The latter case studies provided valuable lessons and analogies that were
applied to the context of GenAI regulation.
• Framework development: Based on the insights gained from the literature review,
comparative analysis, thematic analysis, and case studies, a resilience-focused regula-
tory framework for GenAI was developed. Such a framework emphasizes adaptability,
swift incident response, and recovery mechanisms to mitigate potential harm associ-
ated with GenAI technologies.
Regarding the above-described research methodology, it must be noted that the present
study focused on the regulatory strategies of the EU, China, the US, and the UK to illustrate
divergent global approaches to AI governance. In particular, such regions were selected
Societies 2024, 14, 268 5 of 16

based on their distinct regulatory priorities and methodologies, including risk-based frame-
works (EU), innovation-driven initiatives (the US and UK), and state-controlled policies
(China). Additional countries, such as Russia, Argentina, Japan, and India, are acknowl-
edged for their active AI legislative landscapes; however, they were not included in the
current analysis due to the scope of this study. Indeed, the focus was on selecting world
regions with the highest legislative activity, as identified in the Stanford AI Index 2023,
while also aiming to highlight different approaches to emphasize the current contrasts in
regulatory frameworks and underlying policy objectives.
Moreover, while the present concept paper is focused on the regulation of GenAI, it
is important to notice that GenAI is a subset of the broader category of AI [24]. Indeed,
AI encompasses various forms, including narrow AI, general AI, and super AI. However,
given the recent emergence of GenAI, most global entities have not yet developed specific
regulations exclusively for GenAI [25]. Therefore, the regulatory frameworks discussed in
the current article primarily pertain to AI in general. Such broader focus is necessary to
provide a comprehensive understanding of the current regulatory landscape, as GenAI is
inherently included within the scope of AI regulations. The references to AI regulations
highlight foundational principles and approaches that can be adapted and applied to the
regulation of GenAI. Indeed, as the field of GenAI continues to evolve, it is anticipated
that more specific regulatory measures will be developed to address its unique challenges
and implications. Last, but not least, the present concept paper adopts a cross-sectoral
perspective, recognizing that regulatory challenges and demands may vary significantly
across domains such as education, healthcare, politics, and the military. Thus, the resilience-
based research design outlined in the present article aims to address domain-specific
discrepancies, providing both a robust and adaptable framework that can be applied across
multiple sectors.

3. Comparative Analysis of GenAI and Other Global Challenges

Although the categorization presented in the Introduction Section helps toward the
comprehension of regulatory AI and GenAI laws and frameworks [14], it does not pinpoint
which regulations are the most capable of addressing the specific attributes of GenAI. A
more effective approach might be to compare GenAI technologies with other complex
issues tackled by world governments for several years such as climate change, nuclear
weapons proliferation, and space debris. Such a comparison could rely on the approach
suggested by Roberts [26], which suggests measuring the level of agreement in addressing
the considered problem and the capability to effectively manage it. Indeed, measuring
the people’s consensus and the effectiveness of resolution could help in determining the
gravity of the challenge and the necessary regulations. Regarding the issue of climate
change, multiple global regulations, e.g., the Paris Agreement of 2016 have been signed,
despite the several competing economic interests [27]. As a result, unfortunately, it is
difficult to promote the reduction in emissions. Global politics is mostly aligned regarding
the proliferation of nuclear energy as only nine countries currently have nuclear weapons
and a few governments can actually produce nuclear weapons through the availability
of enriched uranium and related technology [28]. Regarding the issue of space debris,
currently, a few states and space agencies can send rockets into space. However, there
is limited consensus on preventing and/or removing the existing debris, and the several
agencies proceed on their own [29]. Finally, regarding GenAI, several competing interests
are nowadays faced due to the wide economic benefits for companies. Furthermore, it is
difficult to limit the usage of GenAI systems since they are mostly freely available, open
source, and the employed technology to implement them is changing rapidly [5,7,30]. As
a final result, regarding the low consensus toward implementing GenAI regulations and
the current limited control over its applicability, GenAI technologies nowadays present a
particularly complex problem, even potentially worse with respect to the pressing issues of
climate change, the nuclear proliferation of weapons, and the limitation of space debris
Societies 2024, 14, 268 6 of 16

release within space. A summary comparison of GenAI technologies with the analyzed
global challenges is reported in Table 2.
It must be noted that the methodology provided by Roberts [26] was selected in the
context of the present study due to its comprehensive framework, capable of evaluating
complex global challenges. Such a methodology considers both the level of consensus in
addressing a specific problem and the capability to effectively manage it, thereby providing
a thorough approach to regulatory analysis. The latter dual focus is particularly related to
the regulatory landscape of GenAI, which involves multiple stakeholders with different
interests and capabilities. Indeed, by applying the Roberts methodology, in the previous
paragraphs, it was possible to effectively compare the novel challenge of GenAI with other
global challenges such as climate change, nuclear proliferation, and space debris. The
latter comparison both remarks the unique regulatory needs of GenAI technologies and
underscores similarities and differences with the other considered global challenges.

Table 2. Comparison of GenAI technologies with the analyzed global challenges, focusing on
consensus and management capabilities and relying on the approach proposed by Roberts [26]. The
last column of the reported table highlights the distinct regulatory priorities of the analyzed entities,
i.e., the US, UK, China, and the EU.

Level of Management Country/Region

Global Challenge Key Issues
Agreement Capability Highlights
Competing
economic interests, EU: strong
difficulty in environmental
Climate change High Limited
reducing emissions focus; US: varied
despite global state-level policies
agreements [27]
US/UK: focus on
A few countries
global deterrence;
Nuclear have the capability
High Strong China: strategic
proliferation to produce nuclear
military
weapons [28]
considerations
Limited consensus
EU: collaborative
on debris
efforts (e.g., ESA);
prevention,
Space debris Low Strong China: rapid
independent
expansion of space
actions by
programs
agencies [29]
US:
innovation-driven;
Rapidly changing
UK: balanced
GenAI technology, low
Low Limited innovation/safety;
technologies consensus on
China: state
regulations [5,7,30]
control; EU:
citizen welfare

In a similar way to the above-presented issues, GenAI evolves swiftly; exerts a wide-
ranging influence on the economy, society, and governance; demands substantial expertise
for comprehensive understanding; and presents an unpredictable balance of risks and
rewards [5,31]. However, unlike the other presented issues, there is still a lack of consensus:
the multitude of conflicting interests among governments and corporations obstructs the
identification of a shared problem to address, let alone a collective solution [31]. The latter
point is not unexpected, as the potential benefits of AI are so substantial that concerns about
potential risks have not significantly slowed its development [5]. Furthermore, the pace
of AI and GenAI evolution is so rapid that it challenges regulatory bodies to stay abreast.
Indeed, even though there are still numerous tasks where human intuition, creativity, and
emotional intelligence remain unparalleled by AI and GenAI, the advancements in these
technologies over the past decade have been nothing short of remarkable. In just ten years,
AI and GenAI methods have evolved from being relatively underwhelming to outperform-
Societies 2024, 14, 268 7 of 16

ing humans in several domains traditionally dominated by human expertise. Such domains
include, but are not limited to, reading comprehension [32], image recognition [33,34], and
language understanding [35]. The studies reported by Roser [8] provide a comprehensive
overview of these advancements. For instance, AI systems now excel in reading comprehen-
sion tasks, often achieving higher accuracy rates than human counterparts. Similarly, in the
field of image recognition, AI algorithms have reached a level of precision that surpasses
human capabilities, enabling applications ranging from medical diagnostics to autonomous
driving. Language understanding, another critical area, has seen AI models like OpenAI
ChatGPT and its successors demonstrate an impressive ability to generate and comprehend
human-like text. For a more detailed comparison between human and AI capabilities, refer
to Figure 3. The latter figure illustrates the relative performance of AI systems compared to
humans across various tasks, highlighting the areas where AI has not only caught up but
also exceeded human performance.

Figure 3. The figure reports the scores of several AI and GenAI systems on several capabilities
related to human performance. Several capabilities were tested on the models (refer to the legend
on the right of the figure), and the human performance was used as a baseline by setting it to the
value of zero. As a result, AI and GenAI models outperformed humans in several of the analyzed
tasks, for instance, in reading comprehension (18.85), image recognition (16.45), and language
understanding (15.67). On the other hand, they also underperformed compared to humans in many
tasks, such as complex reasoning (−43.12), code generation (−12.64), and math problem solving
(−7.44). The data displayed in the figure were collected from Kiela et al. [36], and figure was
adapted from Roser [8] under the terms of the Creative Commons Attribution License—CC BY 4.0
(https://creativecommons.org/licenses/by/4.0/, accessed on 6 December 2024).

AI and GenAI technologies present multiple opportunities for proper regulatory

interventions that must be taken immediately in the next forthcoming years. Indeed, the
creation of state-of-the-art GenAI models necessitates specialized knowledge, high-end
hardware such as Graphics Processing Units (GPUs), substantial computational power,
and financial resources that could exceed USD 100 million, for instance, in the case of the
development of top-tier GenAI models, such as the OpenAI GPT-4 [37]. However, such a
landscape is evolving and the high expenses associated with the development and operation
of even the most advanced AI and GenAI models are becoming less, thus enfostering
the development of novel models [8,30]. Furthermore, AI models and applications are
increasingly being made available as open source [38]. For instance, meta Llama 3 was
made available to developers in open-source format, but even other models, such as the
Google Gemma model, are also available [39]. Moreover, AI repositories are now leveraged
by AI developers to develop and deploy AI and GenAI models. For instance, the well-
known AI model repository Hugging Face (https://huggingface.co) currently houses
Societies 2024, 14, 268 8 of 16

nearly 800,000 AI models available for download and customization in its transformer
library as of August 2024. The most currently downloaded model, a speech recognition
model, has been downloaded over 292 million times by users within the community [40].

4. How to Safely Apply GenAI Technologies with Resilience?

Differently from the other mentioned problems reported in Section 1, issues with
limited consensus and low control, such as the ones related to GenAI, necessitate prompt
regulations centered on the concept of resilience. Regarding the issue of climate change, soft
laws were applied in the past years where several voluntary agreements were made within
and between many world states, for instance, regarding standards, codes of conduct, and
certifications [27]. Regarding the proliferation of nuclear weapons, historically preventive
agreements and regulations were made to restrict the usage of technologies and actions [28].
Finally, toward the issue of space debris, the civil or criminal liability of space companies
was regulated [29]. However, regarding GenAI, the presence of multiple conflicting interests
between world states and companies and the swift pace of technological advancements
makes it challenging to develop preventive measures and soft regulations. Furthermore,
regulations merely focused on the concept of liability may demand an excessive degree
of control for establishment and enforcement. In contrast, regulations focused on the
concept of resilience may represent a valid alternative, since such regulations could adopt
a permissive approach, thus allowing GenAI technologies to progress and proliferate while
strengthening institutions to formulate principles, standards, procedures, and capabilities to
preemptively address potential issues which may arise due to their usage. Such regulations
may also adopt a retrospective approach, implying they react promptly to issues and
facilitate recovery following any potential incident.
GenAI regulations focused on the concept of resilience necessitate the involvement of
competent and prompt institutions and stakeholders, such as international bodies, national
governments, corporations, civil society, and individuals. Such entities are required to
demonstrate the following seven key resilient features:
• Strong institutions capable of preempting incidents: the incorporation of principles,
standards, procedures, and capabilities that encourage the development, utilization,
and testing of GenAI systems for proactive risk identification and mitigation forms
the bedrock for averting AI-related incidents.
• Swift action to mitigate risks and halt incidents: Events such as intellectual property
theft, the disruption of critical infrastructure, election manipulation through deepfakes,
or other unforeseen occurrences may happen. In such cases, institutions should be
equipped with the necessary authority, procedures, and capabilities to promptly detect
and halt these incidents.
• International collaborations toward unified AI standards: International cooperation is
crucial for developing and enforcing unified standards for GenAI. The latter involves
creating a global framework that ensures consistent regulation, especially in cross-
border applications of AI technologies, to prevent deficiencies in regulations that
could be exploited by bad actors.
• Investments in public awareness and education: Institutions should focus on raising
public awareness about GenAI technologies and their associated risks. By educating
the public and fostering digital literacy, society can better understand GenAI’s im-
pact, which is essential for resilience, particularly in recognizing and mitigating risks
associated with GenAI misuses.
• Fostering innovation with built-in safety mechanisms: Encourage the development of
GenAI systems that inherently include safety features such as AI explainability [41,42],
robustness against adversarial attacks [43], and the ability to deactivate or revert
decisions [44]. This proactive approach ensures that GenAI advancements are aligned
with resilience goals and minimize potential harm from the outset.
• Implementing continuous monitoring and auditing: Establish ongoing monitoring
systems to track the performance and impact of GenAI technologies in real time.
Societies 2024, 14, 268 9 of 16

Regular audits can identify emerging risks early, allowing for swift intervention and
the adjustment of practices to prevent potential harm.
• Bouncing back from risks by reducing harm and providing compensation to victims:
After an incident, it is crucial to ensure that victims, businesses, organizations, and
individuals receive appropriate compensation. Such an approach will maintain trust
in GenAI usage and prevent demands for halting GenAI utilization or implementing
other severe measures.
The actualization of the above-described requirements need the mentioned institutions
to implement certain regulatory measures. At the international level, efforts should focus
on establishing binding GenAI principles and technical standards to create a minimum
threshold for GenAI regulations that can be universally monitored by bodies such as the
G7 and China, or a potential new International Atomic Energy Agency (IAEA) equivalent
for GenAI [45]. In the event of an incident, GenAI risk and incident early warning sys-
tems should be in place to detect and alert the international community about potential
threats. Post-incident, there should be a robust system for tracking and analyzing the
aftermath of GenAI-related incidents to continuously improve global safety standards.
Enhancing international collaboration for unified standards is crucial, as seen in the success
of the Montreal Protocol in phasing out ozone-depleting substances. Investing in public
awareness and education, similar to public health campaigns for disease prevention, can
help society understand GenAI’s impact and recognize risks. Fostering innovation with
built-in safety mechanisms, akin to the international automotive industry’s crash testing
protocols, ensures GenAI advancements align with resilience goals. Continuous monitoring
and auditing, as practiced in environmental regulations to prevent pollution, can identify
emerging risks early [46]. Such measures, combined with swift action to mitigate risks and
halt incidents, will strengthen institutions and ensure a resilient approach to GenAI.
It must be noted that existing multilateral cooperation efforts on AI-related issues
provide a strong foundation for such above-presented initiatives. For instance, the Euro-
pean Commission’s Expert Group on AI [47], the ENISA [48], and the United Nations’ AI
Advisory Body [49] are already working toward establishing robust AI governance frame-
works. Additionally, the Organisation for Economic Co-operation and Development, i.e.,
OECD’s Global Partnership on AI [50] and its Network of Experts [51], as well as initiatives
by the United Nations Interregional Crime and Justice Research Institute (UNICRI) [52]
and the AI for Good initiative by the International Telecommunication Union (ITU) [53],
are fundamental in fostering international collaboration and setting global AI standards.
National governments have a critical role in structuring and enforcing technical stan-
dards within their jurisdictions. This includes licensing GenAI companies, auditing AI
models, establishing post-release controls, conducting red team testing, and implementing
Know Your Customer (KYC) procedures for APIs [54]. Governments should also offer
incentives for companies to exceed these regulatory minimums and invest in developing
more resilient GenAI systems. In response to incidents, there should be an emergency
authority equipped to act swiftly to contain and mitigate the damage. Furthermore, na-
tional governments must establish clear liability and compensation standards to address
the consequences of GenAI incidents. Regarding the available regulations, it must be
noted that the EU’s AI Act outlines specific obligations for member states, including the
designation of national competent authorities for AI oversight [11]. Moreover, Spain was
the first EU country to fulfill the latter obligations by establishing the Spanish Agency for
the Supervision of Artificial Intelligence (AESIA) [55]. Such an agency is responsible for
supervising high-risk AI systems and ensuring compliance with the AI Act.
Enhancing international collaboration for unified standards is crucial, as seen in the
success of the Kyoto Protocol in reducing emissions. Investing in public awareness and
education, similar to national financial literacy programs, can help society understand
AI’s impact and recognize risks. Fostering innovation with built-in safety mechanisms,
akin to the pharmaceutical industry’s clinical trials, ensures AI advancements align with
resilience goals. Continuous monitoring, as practiced in food safety regulations, can identify
Societies 2024, 14, 268 10 of 16

emerging risks early. By taking quick steps to address risks and prevent incidents, such
measures will fortify institutions and promote a robust strategy for GenAI regulation.
Companies must take proactive steps to encourage transparency, share their expertise,
and drive the development of technical standards for GenAI. They should provide data
and models for testing, create watermarks for GenAI-generated content, and develop use
cases for mitigating risks. Internally, companies need to develop controls, processes, and
compliance mechanisms to ensure responsible GenAI practices. If incidents occur, com-
panies should have safety breaks in place, such as requiring human oversight for critical
decisions involving GenAI. To manage the aftermath, companies should contribute to an
insurance fund that can compensate for damages resulting from AI incidents. Enhancing
international collaboration for unified standards is crucial, as seen in the Basel Accords in
banking regulation [56]. Investing in public awareness and education, similar to environ-
mental conservation campaigns, can help society understand AI’s impact and recognize
risks [46]. Fostering innovation with built-in safety mechanisms, akin to the company-level
nuclear industry’s safety protocols, ensures GenAI advancements align with resilience
goals. Continuous monitoring and auditing, as practiced in aviation safety, can identify
emerging risks early. Such actions, together with prompt efforts to manage risks and stop
incidents, will enhance institutions and foster a resilient framework for regulating GenAI.
Civil society has the responsibility to act as a watchdog, identifying best regulatory
practices and advocating for society’s interests in AI governance. This involves conducting
research to identify system risks and developing professional standards for AI practitioners.
When responding to incidents, civil society should work to identify victims, document the
impact of incidents, and push for accountability. In the recovery phase, civil society should
provide independent assessments of the damage caused by AI incidents to ensure fair
and transparent evaluations. Enhancing international collaboration for unified standards
is crucial, as seen in the success of the Stockholm Convention in eliminating persistent
organic pollutants [57]. Investing in public awareness and education, similar to road safety
campaigns, can help society understand AI’s impact and recognize risks. Continuous
monitoring and auditing, as practiced in occupational health and safety, can identify
emerging risks early. Swift actions to mitigate risks and stop incidents, alongside the latter
measures, will bolster institutions and ensure a resilient regulatory framework for GenAI.
Finally, the role of people in shaping the future of AI cannot be understated. Indi-
viduals should have access to AI models, data, and computational resources, and there
should be incentives, such as “bug bounties,” to encourage users to provide feedback and
report potential issues. During the response phase, people can contribute by crowdsourcing
warnings and relevant information about AI-related risks. In the aftermath of incidents,
people should be involved in reporting damages and collecting compensation to ensure
their concerns are addressed and remedied. Enhancing international collaboration for
unified standards is crucial, as seen in the success of the Minamata Convention in reducing
mercury pollution [58]. Investing in public awareness and education, similar to digital
literacy programs, can help society understand GenAI’s impact and recognize risks. Con-
tinuous monitoring and auditing, as practiced in consumer protection regulations, can
identify emerging risks early. Combining the latter measures with timely interventions to
reduce risks and prevent occurrences will reinforce institutions and create a solid approach
to GenAI regulation.
To clearly present the conclusions of the present concept paper, Table 3 finally sum-
marizes the key regulatory measures and actions proposed throughout this research. In
particular, it includes the suggested recommendations and outlines the institutional roles
necessary to ensure a resilient approach to the governance of GenAI. Such a table serves as
a concise summary of the presented conclusions, integrating the multiple aspects of the
resilience-focused regulatory framework discussed in the present article.
Societies 2024, 14, 268 11 of 16

Table 3. Overview of GenAI regulatory measures across institutions, focusing on preemptive actions, risk mitigation, international collaboration, public education,
innovation, monitoring, and compensation.

International Monitoring and Compensation

Institution/Entity Preempt Incidents Mitigate Risks Public Awareness Innovation with Safety
Collaboration Auditing Post-Incident
Establish binding Promote awareness akin Encourage GenAI safety Track and
Implement early Create a unified Monitor global
International bodies GenAI principles to public similar to automotive analyze incidents,
warning systems global framework GenAI practices
and standards health campaigns crash testing protocols improve standards
License GenAI Equip emergency Collaborate on unified
Run public Incentivize resilient Establish liability and
National governments companies, authorities for standards, e.g., Conduct ongoing audits
awareness campaigns GenAI systems compensation systems
audit models swift response Kyoto Protocol
Develop internal Implement safety Share expertise, Engage in awareness Monitor GenAI Contribute to an
Companies and Build GenAI with
controls and breaks, human contribute to and transparency performance, insurance fund
industry inherent safety features
compliance mechanisms oversight standard development initiatives adjust practices for damages
Research and educate Promote responsible Conduct independent Advocate for fair
Advocate for best Identify victims, Push for
Society the public on GenAI use assessments, compensation,
regulatory practices document impacts unified standards
GenAI risks among practitioners identify risks document damage
Provide feedback, Participate in education Encourage transparent
Crowdsource warnings, Support global Report damages,
Individuals report issues (e.g., and awareness and safe Report emerging risks
share information GenAI regulations seek compensation
bug bounties) programs GenAI practices
Societies 2024, 14, 268 12 of 16

5. Future Directions Toward Resilient GenAI Regulations

AI regulations focused on the concept of resilience are already slightly beginning to
surface. The OECD has established an AI incident observatory and categorizes AI systems
based on risk. The above-mentioned EU’s risk classification for AI systems, the US 2023
Executive Order on AI which encourages collaborations with AI firms to report red team
findings, and China’s algorithm registry along with corporate security self-assessments
all incorporate elements of resilience regulations to varying degrees. Indeed, corporations
such as OpenAI, Microsoft, and Google have defined technical standards and have even
made them public, an example being Microsoft’s Responsible AI Standard [59].
The rise in reported AI incidents, as shown in Figure 4, highlights the growing chal-
lenges in ensuring responsible AI deployment and the urgency for resilience measures.
Notable examples of AI incidents are related to deepfake videos [60]. Indeed, the recent mis-
use of deepfake technology has led to significant legal actions in various world countries.
For instance, in the US, the DEEPFAKES Accountability Act was introduced to address
the malicious use of deepfakes [61]. Such legislation aims to hold individuals accountable
for creating and distributing deepfake content with the intent to deceive or harm others.
Moreover, within the UK, a recent notable deepfake case involved the CEO of a UK-based
energy firm who was tricked into transferring EUR 200,000 to a fraudster using deepfake
audio technology to mimic the voice of his superior. Despite the fraudulent nature being
recognized, the company faced challenges in tracing the origins of the deepfake, high-
lighting the need for robust legal frameworks to address such incidents [62]. Although
advised that legal action might be challenging due to existing legislation gaps, such a case
underscored the urgent need for laws to protect individuals from non-consensual deepfake
content [63]. Similarly, the EU Code of Practice on Disinformation calls for measures to deal
with the spread of manipulative deepfake content, particularly in political contexts [64].
Indeed, a recent notable example of the misuse of deepfake technology occurred during
the Russia–Ukraine conflict, where a deepfake video falsely depicting Ukrainian President
Volodymyr Zelenskyy surrendering was circulated. Such a video was quickly debunked,
but it highlighted the potential for deepfakes to be used in disinformation campaigns [65].
Last but not least, AI technology has been used to monitor inmate calls in US prisons,
aiming to detect and prevent criminal activities. Such technology has been employed
to uncover threats, drug smuggling operations, and other illicit activities within the
prison system [66].
There are six further measures that could enhance resilience in the context of GenAI.
Such additional measures were derived from the sources processed during the research,
particularly through the comprehensive analysis of the existing literature, case studies, and
thematic analysis conducted as part of this study, i.e., (1) establish mandatory principles and
baseline technical standards for GenAI, which could be overseen by an organization akin to
the IAEA. (2) Ensure greater accessibility and the provision of resources for civil society and
individuals to utilize data training sets and computational power for model development
and testing. (3) Formulate structured recovery regulations linked to resilience. This could
be the most complex task as it might suggest accountability for incidents. However, in
the absence of explicit liability and compensation norms, GenAI firms might hesitate to
disclose information due to potential future liability concerns. (4) Implement the continuous
scenario-based stress testing of GenAI systems to identify potential vulnerabilities before
they can cause harm. (5) Develop an international framework for AI incident reporting to
standardize the process across borders, ensuring the timely and transparent communication
of risks. (6) Encourage cross-sector collaborations between GenAI developers, regulators,
and industries that heavily rely on GenAI, fostering a shared responsibility in addressing
potential risks and strengthening resilience.
Regarding the previously reported sixth point, the importance of international co-
operation and the establishment of global standards for AI governance was even re-
cently emphasized by the United Nations High-Level Advisory Body on AI [67] which
focused on the following:
Societies 2024, 14, 268 13 of 16

• Ethical AI development: establish ethical guidelines to ensure AI technologies are

developed and used in ways that uphold human rights and promote the social good.
• Inclusive AI: promote inclusivity in AI development, ensuring equitable access to AI
benefits across diverse regions and communities.
• Transparency and accountability: ensure transparency in AI systems and establish ac-
countability mechanisms to prevent misuse and address potential negative consequences.
• Capacity building: focus on building AI expertise and infrastructure in developing
countries, thus enabling their participation in and benefit from AI advancements.
• Alignment with Sustainable Development Goals: align AI governance with the United
Nations Sustainable Development Goals, hence promoting AI applications that con-
tribute to global sustainability objectives.

Figure 4. Annual reported AI incidents and controversies worldwide. The increase in incidents
over the years underscores the growing challenges in managing AI risks. Notable examples include
deepfake videos, such as one falsely depicting Ukrainian President Volodymyr Zelenskyy surren-
dering, and the use of AI to monitor inmate calls in US prisons. Data source: AI Incident Database
via AI Index (2024) [7]. Figure was adapted from Roser [8] under the terms of the Creative Com-
mons Attribution License—CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/, accessed on
6 December 2024).

As a final remark, regulators could enhance consensus and control to shift away from
resilience-focused regulations. The latter one is a logical approach: just as armies strive to
alter power dynamics to achieve victory in warfare, corporations aim to secure monopoly
power. However, modifying such parameters is challenging. It has been forty-four years
since climate change was first discussed at the UN’s inaugural Earth Summit in 1972,
leading to the establishment of binding targets in the 2016 Paris Agreement through the
efforts of advocates and scientists [27]. But the world cannot afford to wait that long: if
GenAI triggers a banking crisis, disrupts critical infrastructure, or inflicts other forms of
widespread societal damage, governments might be compelled to enforce legally binding
authoritarian regulations, irrespective of their practicality or the potential benefits that
might be forfeited. The time to act is now, and the forthcoming AI Safety Summits in France
in 2025 could offer platforms for regulators to more decisively adopt a resilience-oriented
approach, such as the one presented in the present article.

Funding: This research received no external funding.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: No new data were produced.
Societies 2024, 14, 268 14 of 16

Conflicts of Interest: The author declares no conflicts of interest.

References
1. Gupta, P.; Ding, B.; Guan, C.; Ding, D. Generative AI: A systematic review using topic modelling techniques. Data Inf. Manag.
2024, 8, 100066. [CrossRef]
2. Idrisi, M.J.; Geteye, D.; Shanmugasundaram, P. Modeling the Complex Interplay: Dynamics of Job Displacement and Evolution
of Artificial Intelligence in a Socio-Economic Landscape. Int. J. Networked Distrib. Comput. 2024, 12, 185–194. [CrossRef]
3. Gambín, Á.F.; Yazidi, A.; Vasilakos, A.; Haugerud, H.; Djenouri, Y. Deepfakes: Current and future trends. Artif. Intell. Rev. 2024,
57, 64. [CrossRef]
4. Chesterman, S. Good models borrow, great models steal: Intellectual property rights and generative AI. Policy Soc. 2024, puae006.
[CrossRef]
5. Al Naqbi, H.; Bahroun, Z.; Ahmed, V. Enhancing Work Productivity through Generative Artificial Intelligence: A Comprehensive
Literature Review. Sustainability 2024, 16, 1166. [CrossRef]
6. Adigwe, C.S.; Olaniyi, O.O.; Olabanji, S.O.; Okunleye, O.J.; Mayeke, N.R.; Ajayi, S.A. Forecasting the Future: The Interplay of
Artificial Intelligence, Innovation, and Competitiveness and its Effect on the Global Economy. Asian J. Econ. Bus. Account. 2024,
24, 126–146. [CrossRef]
7. Maslej, N.; Fattorini, L.; Perrault, R.; Parli, V.; Reuel, A.; Brynjolfsson, E.; Etchemendy, J.; Ligett, K.; Lyons, T.; Manyika, J.; et al.
The AI Index 2024 Annual Report. AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, Stanford,
CA, April 2024. Available online: https://www.dirittoue.info/?p=4682 (accessed on 6 December 2024).
8. Roser, M. The Brief History of Artificial Intelligence: The world has changed fast—What might be next? Our World Data 2022.
Available online: https://ourworldindata.org/brief-history-of-ai (accessed on 6 December 2024).
9. High-Level Expert Group on Artificial Intelligence. Ethics Guidelines for Trustworthy AI. 2019. Available online:
https://www.aepd.es/sites/default/files/2019-12/ai-definition.pdf (accessed on 6 December 2024).
10. European Union Agency for Cybersecurity (ENISA). AI Cybersecurity Challenges—Threat Landscape for Artificial Intelligence.
2021. Available online: https://protecciondata.es/wp-content/uploads/2021/06/Panorama-de-amenazas-de-ENISA-sobre-
inteligencia-artificial-informe-de-2020.pdf (accessed on 6 December 2024).
11. The European Commission. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024. Off. J.
Eur. Union 2024. Available online: http://data.europa.eu/eli/reg/2024/1689/oj (accessed on 6 December 2024).
12. Wörsdörfer, M. Biden’s Executive Order on AI and the E.U.’s AI Act: A Comparative Computer-Ethical Analysis. Philos. Technol.
2024, 37, 74.[CrossRef]
13. Davison, N. A Legal Perspective: Autonomous Weapon Systems under International Humanitarian Law. UNODA Occas.
Pap. 2017, 30, 5–18. Available online: https://www.icrc.org/en/document/autonomous-weapon-systems-under-international-
humanitarian-law (accessed on 27 November 2022).
14. Bryer, A.R. Understanding Regulation: Theory, Strategy, and Practice. Account. Eur. 2013, 10, 279–282. [CrossRef]
15. The White House. Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. 2023.
Available online: https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-
secure-and-trustworthy-development-and-use-of-artificial-intelligence/ (accessed on 6 December 2024).
16. Illinois General Assembly. Artificial Intelligence Video Interview Act. 2019. Available online: https://www.ilga.gov/legislation
/publicacts/fulltext.asp?Name=101-0260 (accessed on 6 December 2024).
17. Government of the United Kingdom—Department for Science, Innovation & Technology. Implementing the UK’s AI Regulatory
Principles: Initial Guidance for Regulators. 2024. Available online: https://www.gov.uk/government/publications/implement
ing-the-uks-ai-regulatory-principles-initial-guidance-for-regulators/102aa401-60f6-46e8-95dc-b48150eba7dd (accessed on 6
December 2024).
18. Roberts, H.; Babuta, A.; Morley, J.; Thomas, C.; Taddeo, M.; Floridi, L. Artificial intelligence regulation in the United Kingdom: A
path to good governance and global leadership? Internet Policy Rev. 2023, 12, 1–31. [CrossRef]
19. Cyberspace Administration of China. Regulations on the Administration of Internet Information Service Recommendation
Algorithms. 2022. Available online: https://www.gov.cn/zhengce/zhengceku/2022-01/04/content_5666429.htm (accessed on
6 December 2024).
20. Cyberspace Administration of China. Provisions on the Administration of Deep Synthesis in Internet-Based Information Services.
2023. Available online: http://www.cac.gov.cn/2022-01/28/c_1644970458520968.htm (accessed on 6 December 2024).
21. Cyberspace Administration of China. Interim Measures for the Management of Generative Artificial Intelligence Services. 2023.
Available online: https://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm (accessed on 6 December 2024).
22. Franks, E.; Lee, B.; Xu, H. Report: China’s New AI Regulations. Glob. Priv. Law Rev. 2024, 5, 43–49. [CrossRef]
23. Fung, P.; Etienne, H. Confucius, cyberpunk and Mr. Science: Comparing AI ethics principles between China and the EU. AI Ethics
2023, 3, 505–511. [CrossRef] [PubMed]
24. Damar, M.; Özen, A.; Çakmak, Ü.E.; Özoğuz, E.; Safa Erenay, F. Super AI, Generative AI, Narrow AI and Chatbots: An Assessment
of Artificial Intelligence Technologies for The Public Sector and Public Administration. J. AI 2024, 8, 83–106. [CrossRef]
25. Huang, K.; Joshi, A.; Dun, S.; Hamilton, N. AI Regulations. In Generative AI Security. Future of Business and Finance; Huang, K.,
Wang, Y., Goertzel, B., Li, Y., Wright, S., Ponnapalli, J., Eds; Springer: Cham, Switzerland, 2024. [CrossRef]
Societies 2024, 14, 268 15 of 16

26. Roberts, N. Wicked Problems and Network Approaches to Resolution. Int. Public Manag. Rev. 2000, 1, 1–19. Available online:
https://ipmr.net/index.php/ipmr/article/view/175 (accessed on 6 December 2024).
27. Bodansky, D. The legal character of the Paris agreement. Rev. Eur. Comp. Int. Environ. Law 2016, 25, 142–150. [CrossRef]
28. Ghoshal, D. Historical and Contemporary Missile Development: Nuclear Weapon States, Regional Powers and Other Powers. In
Role of Ballistic and Cruise Missiles in International Security; Palgrave Macmillan: Cham, Switzerland, 2023; pp. 69–143. [CrossRef]
29. Klimburg-Witjes, N. A Rocket to Protect? Sociotechnical Imaginaries of Strategic Autonomy in Controversies About the European
Rocket Program. Geopolitics 2023, 29, 821–848. [CrossRef]
30. Albayrak Ünal, Ö.; Erkayman, B.; Usanmaz, B. Applications of Artificial Intelligence in Inventory Management: A Systematic
Review of the Literature. Arch. Comput. Methods Eng. 2023, 30, 2605–2625. [CrossRef]
31. Saheb, T.; Saheb, T. Topical review of artificial intelligence national policies: A mixed method analysis. Technol. Soc. 2023,
74, 102316. [CrossRef]
32. Xiao, C.; Xu, S.X.; Zhang, K.; Wang, Y.; Xia, L. Evaluating Reading Comprehension Exercises Generated by LLMs: A Showcase of
ChatGPT in Education Applications. In Proceedings of the 18th Workshop on Innovative Use of NLP for Building Educational
Applications (BEA 2023), Toronto, Canada, 13 July 2023; pp. 610–625. [CrossRef]
33. Bodini, M. Will the Machine Like Your Image? Automatic Assessment of Beauty in Images with Machine Learning Techniques.
Inventions 2019, 4, 34. [CrossRef]
34. Bodini, M. A Review of Facial Landmark Extraction in 2D Images and Videos Using Deep Learning. Big Data Cogn. Comput. 2019,
3, 14. [CrossRef]
35. Karanikolas, N.; Manga, E.; Samaridi, N.; Tousidou, E.; Vassilakopoulos, M. Large Language Models versus Natural Language
Understanding and Generation. In Proceedings of the 27th Pan-Hellenic Conference on Progress in Computing and Informatics
(PCI ’23), Lamia, Greece, 24–26 November 2023; pp. 278–290. [CrossRef]
36. Kiela, D.; Thrush, T.; Ethayarajh, K.; Singh, A. Plotting Progress in AI. Contextual AI Blog 2023. Available online:
https://contextual.ai/blog/plotting-progress (accessed on 6 December 2024).
37. Telenti, A.; Auli, M.; Hie, B.L.; Maher, C.; Saria, S.; Ioannidis, J.P.A. Large language models for science and medicine. Eur. J. Clin.
Investig. 2024, 54, e14183. [CrossRef] [PubMed]
38. Hagg, A.; Kirschner, K.N. Open-Source Machine Learning in Computational Chemistry. J. Chem. Inf. Model. 2023, 63, 4505–4532.
[CrossRef] [PubMed]
39. Chang, Y.; Wang, X.; Wang, J.; Wu, Y.; Yang, L.; Zhu, K.; Chen, H.; Yi, X.; Wang, C.; Wang, Y.; et al. A Survey on Evaluation of
Large Language Models. ACM Trans. Intell. Syst. Technol. 2024, 15, 39. [CrossRef]
40. Gong, Y.; Chung, Y.A.; Glass, J. Ast: Audio spectrogram transformer. In Proceedings of the Annual Conference of the International
Speech Communication Association, INTERSPEECH, Brno, Czechia, 30 August–3 September 2021; pp. 571–575. [CrossRef]
41. Bodini, M.; Rivolta, M.W.; Sassi, R. Interpretability Analysis of Machine Learning Algorithms in the Detection of ST-Elevation
Myocardial Infarction. Comput. Cardiol. 2020, 47, 1–4. [CrossRef]
42. Bodini, M.; Rivolta, M.W.; Sassi, R. Opening the Black Box: Interpretability of Machine Learning Algorithms in Electrocardiogra-
phy. Philos. Trans. R. Soc. A 2021, 379, 20200253. [CrossRef] [PubMed]
43. Dong, H.; Dong, J.; Wan, S.; Yuan, S.; Guan, Z. Transferable Adversarial Distribution Learning: Query-Efficient Adversarial
Attack Against Large Language Models. Comput. Secur. 2023, 135, 103482. [CrossRef]
44. Chiang, C.-W.; Lu, Z.; Li, Z.; Yin, M. Enhancing AI-Assisted Group Decision Making through LLM-Powered Devil’s Advocate. In
Proceedings of the 29th International Conference on Intelligent User Interfaces (IUI ’24), Greenville, SC, USA, 18–21 March 2024;
pp. 103–119. [CrossRef]
45. De Castro, B. L.; Gracia, F. J.; Peiró, J. M.; Pietrantoni, L.; Hernández, A. Testing the Validity of the International Atomic Energy
Agency (IAEA) Safety Culture Model. Accid. Anal. Prev. 2013, 60, 231–244. [CrossRef] [PubMed]
46. Bodini, M. Charting the Future of Conservation in Arizona: Innovative Strategies for Preserving Its Natural Resources. Conserva-
tion 2024, 4, 402–434.[CrossRef]
47. European Commission. Expert Group on AI. Available online: https://digital-strategy.ec.europa.eu/en/policies/expert-group-ai
(accessed on 6 December 2024).
48. European Union Agency for Cybersecurity (ENISA). Available online: https://www.enisa.europa.eu/ (accessed on
6 December 2024).
49. United Nations. AI Advisory Body. Available online: https://www.un.org/techenvoy/ai-advisory-body (accessed on
6 December 2024).
50. OECD. Global Partnership on AI. Available online: https://www.oecd.org/en/about/programmes/global-partnership-on-
artificial-intelligence.html (accessed on 6 December 2024).
51. OECD. Network of Experts. Available online: https://oecd.ai/en/network-of-experts (accessed on 6 December 2024).
52. United Nations Interregional Crime and Justice Research Institute (UNICRI). Accessed on: https://unicri.it/topics/ai_robotics/
(accessed on 6 December 2024).
53. International Telecommunication Union (ITU). AI for Good. Available online: https://aiforgood.itu.int/ (accessed on
6 December 2024).
54. Kapsoulis, N.; Psychas, A.; Palaiokrassas, G.; Marinakis, A.; Litke, A.; Varvarigou, T. Know Your Customer (KYC) Implementation
with Smart Contracts on a Privacy-Oriented Decentralized Architecture. Future Internet 2020, 12, 41. [CrossRef]
Societies 2024, 14, 268 16 of 16

55. Spanish Agency for the Supervision of Artificial Intelligence (AESIA). Available online: https://www.lamoncloa.gob.es/lang/en
/gobierno/news/Paginas/2024/20240619-ai-oversight-agency.aspx (accessed on 6 December 2024).
56. ElBannan, M. A. The Financial Crisis, Basel Accords and Bank Regulations: An Overview. Int. J. Account. Financ. Report. 2017, 7,
225–275. [CrossRef]
57. Hagen, P. E.; Walls, M. P. The Stockholm Convention on Persistent Organic Pollutants. Nat. Resour. Environ. 2005, 19, 49–52.
Available online: https://www.jstor.org/stable/40924611 (accessed on 6 December 2024).
58. Evers, D.C.; Keane, S.E.; Basu, N.; Buck, D. Evaluating the Effectiveness of the Minamata Convention on Mercury: Principles and
Recommendations for Next Steps. Sci. Total Environ. 2016, 569, 888–903. [CrossRef] [PubMed]
59. Camilleri, M.A. Artificial intelligence governance: Ethical considerations and implications for social responsibility. Expert Syst.
2024, 41, e13406. [CrossRef]
60. De Rancourt-Raymond, A.; Smaili, N. The unethical use of deepfakes. J. Financ. Crime 2023, 30, 1066–1077. [CrossRef]
61. DEEPFAKES Accountability Act (H.R. 3230). 116th Congress (2019–2020). Available online: https://www.congress.gov/bill/116th-
congress/house-bill/3230 (accessed on 6 December 2024).
62. Quirk, C. The High Stakes of Deepfakes: The Growing Necessity of Federal Legislation to Regulate This Rapidly Evolving
Technology. Princet. Leg. J. 2023. Available online: https://legaljournal.princeton.edu/the-high-stakes-of-deepfakes-the-growing-
necessity-of-federal-legislation-to-regulate-this-rapidly-evolving-technology/ (accessed on 6 December 2024).
63. HyperVerge. Are Deepfakes Illegal? 2023. Available online: https://hyperverge.co/blog/are-deepfakes-illegal/ (accessed on
6 December 2024).
64. Borz, G.; De Francesco, F.; Montgomerie, T.L.; Bellis, M.P. The EU soft regulation of digital campaigning: Regulatory effectiveness
through platform compliance to the code of practice on disinformation. Policy Stud. 2024, 45, 709–729. [CrossRef]
65. Sky News. Ukraine War: Deepfake Video of Zelenskyy Telling Ukrainians to ’Lay Down Arms’ Debunked. 2022. Available
online: https://news.sky.com/story/ukraine-war-deepfake-video-of-zelenskyy-telling-ukrainians-to-lay-down-arms-debunked-
12567789 (accessed on 6 December 2024).
66. ABC News. US Prisons and Jails Using AI to Mass-Monitor Millions of Inmate Calls. 2019. Available online: https://abcnews.go
.com/Technology/us-prisons-jails-ai-mass-monitor-millions-inmate/story?id=66370244 (accessed on 6 December 2024).
67. United Nations. Governing AI for Humanity: Final Report of the UN High-Level Advisory Body on Artificial Intelligence. 2024.
Available online: https://www.un.org/sites/un2.un.org/files/governing_ai_for_humanity_final_report_en.pdf (accessed on
6 December 2024).

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.
Reproduced with permission of copyright owner. Further reproduction
prohibited without permission.