The Federal Trade Commission (FTC) is reportedly in the initial stage of investigating OpenAI regarding allegations that its flagship conversational AI model, ChatGPT, made false, misleading, disparaging, and harmful statements about individuals.
While it is unlikely that this investigation will result in an immediate crackdown, it signifies that the FTC is taking more than just a cautionary stance towards potential violations within the AI industry.
According to The Washington Post, which claims to have access to a 20-page letter sent to OpenAI, the FTC has requested information on complaints related to the issue.
As investigations conducted by the FTC are nonpublic, the agency declined to comment on the matter.
In February, the FTC established the Office of Technology to address deceptive practices in the technology sector.
Shortly thereafter, the FTC warned companies making claims about AI that they are subject to the same truth requirements as any other entity.
In a public announcement, the FTC emphasized the need for companies to ensure the accuracy of their AI-related claims to avoid regulatory actions.
Although the reported letter is not the first instance of the FTC engaging with AI-related issues, it suggests that OpenAI, as the current unrivaled leader in the field, must be prepared to provide justifications for its actions.
It is important to note that such investigations are not initiated without cause. The FTC does not randomly select targets based on suspicion alone.
Typically, a lawsuit or formal complaint brings a company’s practices to the FTC’s attention, indicating potential violations of regulations.
For example, if a person sues a supplement company claiming that their product caused illness, the FTC may launch an investigation based on evidence that the company made false claims about the product’s side effects.
In this case, it is highly probable that a lawsuit prompted the investigation. An Australian mayor allegedly complained to OpenAI after ChatGPT purportedly stated that he had been involved in bribery and sentenced to prison, among other defamatory claims.
Although the jurisdiction is incorrect, this case likely represents one of many similar incidents.The publication of such damaging information could potentially amount to defamation, libel, or reputational damage, as referred to in the FTC’s letter to OpenAI.
It is most likely ChatGPT that is at the center of the investigation, as it is OpenAI’s most prominent public product capable of causing such harm.
OpenAI’s more recent and restricted offerings, such as GPT-4 and other APIs, are less likely to be considered in this investigation due to their limited access and recency.
However, establishing liability in this situation is not straightforward. The technical aspects alone raise questions about whether this constitutes publishing, speech, or simply private communication.
All these factors would need to be proven. Nevertheless, it is reasonable to request that a company provide an explanation.
Making an occasional mistake is one thing, but systematically and invisibly fabricating details about individuals on a large scale without disclosure is a different matter.