1

Giles Buxton

English 101

Professor Saki Gejo

10/8/24

AI Security Proposal

Over the past few years, artificial intelligence has been rapidly implemented into

countless businesses around the world. These companies see artificial intelligence as the next big

step to evolve their businesses as it can do mundane or slow tasks efficiently and consistently

allowing resources and labor to be focused on other parts of the business. However, many

experts and users have become increasingly concerned about the consequences of the

implementation. Having AI have access to personal user information and important data makes

them a target for malicious cyber attacks. While not much research has been conducted on the

security of artificial intelligence, researchers have already found that artificial intelligence is

susceptible to data poisoning. This is a large security threat as if an AI is trained on poisoned

data it can give an attacker access to the program and in turn potentially allow them into

confidential data and/or private user information. This brings up the question of how artificial

intelligence should be regulated in businesses in order to protect data and users alike. This is

currently a huge gray zone as outrighting banning the use of AI in businesses could be sacrificing

potential economic growth while on the other hand, letting AI rampant in businesses could put

secure information at risk. In order to solve this problem, government officials need to step in

and develop regulations on the use of artificial intelligence in businesses. Currently leaving AI

uncontrolled is putting citizens at risk and a simple ban on AI may cause more harm than good
 2

especially for the economy. What governments across the world need to do is fund research on

the security of artificial intelligence as well as how AI stores and processes data. Once this is

completed accurate regulations can be placed on artificial intelligence dictating where it can be

used in businesses and what it can and cannot do in relation to data analysis.

However, many claim that artificial intelligence is extremely beneficial to society as a

whole so instead of putting strict regulations on it, it should be encouraged in order to maximize

the pros it offers. One of the largest claims being made is that artificial intelligence being

implemented into businesses could significantly boost the economy Manyika, Smit, Woetzel,

Birshan, Russel, and Russel (2021), explain that in OECD economies businesses make up 72

percent of the economies total GDP where the rest counts for government, non-profit, and

household incomes from real estate. With so much of the national GDP reliant on businesses,

people fear that if we regulate how AI can be used in these businesses, we could be sacrificing

unprecedented economic growth, perhaps even falling behind other countries who decide to

leave it unregulated. Not only do people claim that AI could lead to economic growth, but that it

could also allow us to allocate resources more efficiently. The U.S. Department of Labor states “

AI can positively augment work by replacing and automating repetitive tasks or assisting with

routine decisions, which may reduce the burden on workers and allow them to better perform

other responsibilities.” (U.S. Department of Labor). By limiting AI use, it is possible that it may

affect workers as they will have to spend more time doing mundane tasks that do not require

their expertise in turn wasting their time and leading to lower productivity. The combination of

these two major points have led many to believe that despite possible risks that AI will have in

businesses, the upsides significantly outweigh them and instead of wasting time making

regulations, we should instead focus on implementing and improving AI in the workforce.

While these are sound claims, AI is much too harmful when misused to allow it to be

regulated by businesses' own discretion. There are significant flaws that are apparent when it

comes to the security of artificial intelligence and the possible information that it handles. The

primary vulnerability is its susceptibility to data poisoning. As described by Nihad Hassan

(2024), data poisoning is where artificial intelligence data sets are corrupted with false or bugged

information while being trained leading to the program becoming corrupted. Though this sounds

fairly minor and may only lead to a few bugs on the surface level, this leads to many more major

issues that attackers can take advantage of in order to gain access to sensitive data. One found

example of this is that when graph neural networks or a GNN (an artificial neural network used

for processing data) is trained on poisoned data, it can give an attacker a trigger in order to

receive a backdoor into the program. This is fairly undetectable since the trigger isn’t commonly

used in regular data sets making it hard for the administrators to even know that the program has

been poisoned (Zhang, Dai, Lin, & Wang, 2024). With AI being vulnerable to such attacks, we

as a country should not aim to overlook the possibility of malicious attacks on AI in order to

capitalize on possible economic gains even more so if we are leaving the issue to businesses to

do what they see fit with the technology.

Many businesses have made it abundantly clear that regardless of the possible risks, they

will still strive towards mainstream use of the systems internally. Kerkez Nina wrote, “Numerous

banks in the market are exploring the option of utilizing new technologies as a labor-saving

innovation due to the extremely high costs associated with compliance staff.” (Nina, 2020).

Businesses will always aim towards maximizing their profits while keeping their costs low and

they see AI as the perfect new venture in order to do this. Not only does this present issues
 4

towards job security, but it also leads to the overlook of security threats as these companies do

not understand or need to understand the risks involved with AI.

That is why I propose that national governments step in and regulate this advanced

technology in order to not only ensure the safety of sensitive information, but also to allow the

safe use of AI in businesses to boost the economy. In order to best do this, governments will

need to fund research on security threats on AI, how AI handles data, and then implement

regulations on its use commercially.

Temp Links / references

https://www.dol.gov/general/ai-principles#:~:text=AI%20can%20positively%20augment

%20work,to%20better%20perform%20other%20responsibilities

Manyika, J., Birshan, M., Smit, S., Woetzel, J., Russell, K., & Purcell, L. (2021, May 31).

A new look at how corporations impact the economy and households | McKinsey.

Www.mckinsey.com.

Hassan, N. (2024, May). What is data poisoning (AI poisoning) and how does it work?

Enterprise AI. https://www.techtarget.com/searchenterpriseai/definition/data-poisoning-

AI-poisoning

Zhang, Z., Lin, M., Dai, E., & Wang, S. (2024). Rethinking graph backdoor attacks: A

distribution-preserving perspective. In Proceedings of the 30th ACM SIGKDD

Conference on Knowledge Discovery and Data Mining (KDD '24) (pp. 4386–4397).

Association for Computing Machinery. https://doi.org/10.1145/3637528.3671910

Kerkez, N. (2020). Artificial Intelligence and Machine Learning Can Repurpose Humans,

Not Replace Them. American Bankers Association.ABA Banking Journal, 112(6), 30-32.

http://login.ezproxy1.lib.asu.edu/login?url=https://www.proquest.com/scholarly-

journals/artificial-inteligence-machine-learning-can/docview/2460794899/se-2