The recent buzz in business tech is artificial intelligence, or AI. The advantages of AI for businesses are great, including automated workflow, increased speed of routine tasks, and improved efficiency in the office. However, there are a number of hazards that come with AI, including privacy issues, security, and, as was just learned this week, basic questions of reliability and truth.
In a case filed in the Southern District of New York, an attorney presented a motion to dismiss that had been created with the assistance of the popular AI ChatGPT. To the attorney’s dismay and the judge’s fury, the brief was filled with “bogus judicial decisions, bogus quotes, and bogus internal citations.” Situations like these lend impetus to the FTC’s recent decisions to regulate AI technology and AI-powered products.
Federal AI Regulations
There are laws in some U.S. states regulating AI. These laws have no coherent structure and are variously aimed at requiring social media platforms to be transparent about code violations (California), requiring employers to disclose the use of AI in hiring and interviewing practices (Illinois), and studying the use of AI-supported systems in government (New York, Washington, Rhode Island).
The Federal Trade Commission (FTC) has primarily been focused on false, misleading, and unsubstantiated claims about the use of AI in products. Specifically, the FTC is concerned with:
- Exaggerated or false claims: Claims must be supported by performance. A seller cannot claim their AI product can predict human behavior or guarantee results — or, as the hapless New York attorney discovered, cite case law.
- Claims that an AI product is better than a non-AI product: Any comparison must be backed up by facts, and if the facts don’t bear out the results, the claim cannot be made.
- Risks that are downplayed or disregarded: If there are risks involved with the use of the AI product, such as bias or a known failure rate, the seller must make potential buyers aware of the possible failure and risks. The seller cannot pass this off on the developer or coder—all risks must be accounted for before marketing.
- Inaccurate claims that a product uses AI: Just saying a product is “AI-powered” or was developed with AI does not make it AI-enabled, and the FTC is checking all such claims.
The risk of bias in AI use needs to be mentioned here. Generative AI — that is, artificial intelligence generates output based upon existing data — can only create unbiased output if the data it is given are unbiased. An AI “learns” based upon its input. For instance, a hiring AI “taught” to rank resumes that were primarily male applicants will eventually return only male candidates, even if that was not intended.
For this reason, the FTC, the Civil Rights Division of the Department of Justice, the U.S. Equal Employment Opportunity Commission (EEOC), and the Consumer Financial Protection Bureau joined together in a statement to “uphold America’s commitment to the core principles of fairness, equality, and justice as emerging automated systems, including those sometimes marketed as ‘artificial intelligence’ or ‘AI,’ become increasingly common in our daily lives — impacting civil rights, fair competition, consumer protection, and equal opportunity.”
State AI Regulation and Legislation
State legislation creating AI task forces and commissions have increased during the last three years, and states have boosted their efforts to regulate AI products and services. The majority of these regulatory efforts have failed, but the states have continued to work towards restricting unlimited AI growth.
State efforts have been focused on ways to bring AI into the government sphere, including:
- Predictive policing technology
- Facial recognition technology, which is already in use in the private sector
- Consumer rights, based on FTC guidance
- Employment rights and protections
- Healthcare issues
Some bills enacted or in progress address potential problems before they become issues. For instance, New York City’s “AI Law” regulates the use of automated employment decision-making tools in hiring. California is considering AB331, which would require a two-step “deployer” and “developer” for any automatic decision-making platform to carry out an initial assessment of the impact of the tool.
What Businesses Should Consider for the AI Future
Artificial intelligence is here, and the genie won’t be going back into the bottle. Companies that create or use AI need to consider the best way to implement AI within their business and how to avoid the potential risks of using it.
Understand What AI is Within Your Business
The best-known AI today is ChatGPT. Also known as OpenAI or GPT-3, this system is considered the most powerful AI on the market today. It is a “generative AI,” meaning it creates text, images, code, and new data from previously inputted data. Generative AI can identify patterns in data and “predict” new outcomes based on them.
Companies use AI to streamline their decision-making process, especially when large amounts of similar data need to be analyzed. For instance, when sorting through resumes, an AI can be “taught” to look for certain keywords or word strings and flag resumes with those words for later review.
Companies should verify when they are using AI decision-making tools and what impact this has on their operations.
Conduct Regular Risk Assessments on AI Technologies
Use FTC guidance to assess the potential risks and benefits of AI tools. Consider where bias, lack of transparency, or uncertainty about effectiveness may cause possible problems. The FTC recommends that companies using AI consider creating their own internal “AI commission” for analyzing the costs and benefits of new technologies before switching to the newest exciting platform.
Companies should also be sure to conduct assessments or audits on their AI and other technologies regularly. Things change rapidly in the tech world, so don’t become complacent and think that fixing something once means it is cured for all time.
Responsibility for Compliance Must Go Both Directions
Compliance with FTC, state, and organizational AI regulations must be enforced at all levels of the company. Although upper management may draft policies regarding accountability and risk management, and employees may be tasked with following the policies, all tiers of the company must strictly follow the policies.
Policies may vary depending on the nature of the company, but should include:
- Transparency with training, algorithm development, and data set creation
- Accuracy and fact-checking
- Fairness and social impact
- Bias training and reporting
All employees should be empowered to report a possible breach in the policy to the appropriate department or manager.
Create or Designate a Risk Management Officer
In an AI-reliant company, it might be necessary to assign an individual whose job is solely to oversee AI policies and procedures. In a company in which AI is less critical or a smaller part of the company systems, the policy oversight could be designated to the existing risk management or compliance officer.
In either case, the risk management and compliance program must become as much a part of corporate culture as any other safety and regulatory compliance program. Management and employees must become aware of the limitations and risks of AI-decision making tools and the proper way to avoid falling into the traps that got a New York attorney an Order to Show Cause.
Global Impact of AI Regulations
Companies that operate overseas should also know that the European Union (EU) has proposed the world’s first comprehensive AI law, the EU Artificial Intelligence Act. The AI Act is set to go before the European Parliament for voting later this year. It would greatly restrict AI applications considered to create an “unacceptable risk” of impacting human lives, such as the type of social scoring system used in China. Employment-ranking AI tools would receive a “high-risk application” tag and be heavily regulated.
Other countries that are not signatory to the EU are also considering strict regulation of ChatGPT and other generative AI tools and applications. It is in all companies’ best interests to self-regulate their use of AIs before the U.S. government or other nations’ governments do it for them.
No matter how your company uses artificial intelligence, whether as a decision-making tool or an assistant for database management, AI and regulatory decisions will affect your business in the years to come. Your best move is to take control of your AI system and make it work for you. Understanding the risks and benefits of AI and what it can and cannot do are the first steps to securing your AI future.
The United States’ Approach to AI Regulation: Key Considerations for Companies JD Supra, May 23, 2023
A lawyer used ChatGPT and now has to answer for its ‘bogus’ citations The Verge, May 27, 2023
Legislation Related to Artificial Intelligence National Conference of State Legislators, August 26, 2022
Keep your AI claims in check FTC Business Blog, February 27, 2023
FTC Chair Khan and Officials from DOJ, CFPB and EEOC Release Joint Statement on AI FTC Website, April 25, 2023
The Artificial Intelligence Act The Artificial Intelligence Act, accessed May 28, 2023