It was only a matter of time before the Federal Trade Commission (FTC) initiated a law enforcement investigation focusing on artificial intelligence (AI). It is not an exaggeration to say that the FTC has been hinting at, and sometimes outright foreshadowing, such actions for years. The frequency of these communications has only increased over the past few months as generative AI technology dominated headlines.
The FTC has launched an investigation of OpenAI Inc.’s generative AI technology, including whether OpenAI has engaged in “unfair or deceptive privacy or data security practices” or “unfair or deceptive practices relating to risks of harm to consumers, including reputational harm.” It is fairly rare for FTC consumer protection investigations to leak, but last week the investigation was widely reported in the media and the actual Civil Investigation Demand (CID) itself also leaked and can be seen here.
In the 20-page CID, the FTC sets forth a number of interrogatories and document requests focusing on how OpenAI has built, tested and advertised its large language model products. With respect to the training data sets, the FTC asks how OpenAI obtained the data it used to train its technology, how it vets and categorizes data, and how it filters or anonymizes personal information in the data sets. OpenAI will have to explain how it trains its models, including fine-tuning, retraining and reinforcement learning through human feedback. The FTC is asking how OpenAI handles what the industry generally recognizes as an inevitable problem of generative AI: hallucinations. The investigation will also look into the preventive and remedial measures OpenAI employs to handle the disclosure of personal information and the creation of false, misleading or disparaging statements about individuals.
The FTC is asking the same questions of OpenAI that enterprise users of ChatGPT should be considering. For businesses looking to build their own technology on top of OpenAI’s generative pre-trained transformer models, they should not rely on OpenAI to build in all the protections necessary to comply with applicable law. Instead, enterprise users should be conducting their own due diligence on their specific usage and implementation of generative AI technology.
Given the FTC’s repeated focus on a wide range of potentially deceptive or unfair features of generative AI, it is safe to assume that this is not the only AI investigation in the works and that there are many practices that could be the subject of extensive FTC inquiry. The agency continues to broadly interpret its unfairness authority, and that will likely continue as it investigates more consumer protection concerns regarding AI.