OpenAI, a leading artificial intelligence (AI) company, is reportedly facing an investigation by the U.S. Federal Trade Commission (FTC) over potentially deceptive business practices. The investigation revolves around concerns over OpenAI’s language models and their ability to generate false or disparaging statements about individuals, as well as the company’s privacy and data security practices.
According to a regulatory document recently made public, the FTC has issued a civil subpoena to OpenAI to gather information on these matters. The release of the subpoena comes after an AI ethics group had previously requested the FTC to investigate the startup.
The FTC’s investigation focuses on two main areas. Firstly, the agency is looking into whether OpenAI has engaged in unfair or deceptive practices that could potentially harm consumers, including reputational harm. Specifically, the FTC is interested in understanding what steps OpenAI is taking to prevent its language models from generating false or disparaging statements about individuals.
Secondly, the FTC is examining whether OpenAI has been involved in unfair or deceptive privacy or data security practices. While the specific practices under review have not been disclosed, the subpoena includes a request for information regarding a cybersecurity issue discovered in OpenAI’s ChatGPT earlier this year.
In March, OpenAI addressed an issue that allowed certain ChatGPT users to access a limited amount of data from other users’ chat histories. Additionally, partial payment information belonging to some paying customers was briefly accessible via the chatbot. The FTC is seeking further information on this incident, as well as any other data security vulnerabilities that may have affected OpenAI’s large language models.
Prompt injection attacks, wherein hackers manipulate a language model’s output using malicious prompts, are also of interest to the FTC. The agency has requested information on whether OpenAI has been targeted by such cyberattacks.
In addition to these specific concerns, the FTC subpoena covers various other aspects, including the disclosure of earnings data and details on OpenAI’s AI model training methods. The agency is also seeking information on the company’s risk assessment and safety evaluation procedures before releasing a new Large Language Model product.
This investigation by the FTC follows a wave of litigation targeting not only OpenAI but also other generative AI providers. One class-action lawsuit accuses OpenAI of using personal information to train its language models. Additionally, OpenAI and Microsoft Corp. were recently sued for allegedly breaching open-source software project licensing terms to develop Copilot, an AI-powered code completion tool.
As the investigation unfolds, OpenAI and other AI companies will likely face scrutiny over their data practices and how they address potential risks and vulnerabilities associated with their language models. It remains to be seen what the outcome of the FTC’s investigation will be and what implications it may have for the AI industry as a whole.
