OpenAI and Anthropic face a U.S. regulatory investigation over security incidents, including a recent string of controversies involving AI agents carrying out hacking attacks.
The Federal Trade Commission sent civil investigative demands to OpenAI and Anthropic on the 14th local time, requiring the companies to turn over records related to the security incidents, the U.S. online outlet Axios reported. A civil investigative demand is an administrative procedure comparable to a court subpoena, used by authorities to obtain internal corporate documents. In the 50-page demand, the FTC sought materials on recent security incidents and related safeguards, focusing on possible consumer protection violations, according to the report. The agency also plans to summon executives from each company soon to hear testimony on how their AI products affect U.S. consumers.
The Financial Times reported on Sept. 30 that the FTC planned to demand documents from major companies including OpenAI and Anthropic over potential harms from AI products, and to secure testimony from top executives.
The moves signal that U.S. authorities have begun responding in earnest to a series of incidents in which AI models displayed aggressive and anomalous behavior. The most prominent case came in July, when OpenAI's "rogue agents" broke out of a sandboxed environment on their own and attacked the systems of Hugging Face, an outside organization. OpenAI Chief Executive Sam Altman described it as the most serious case identified to date.
More recently, an OpenAI AI agent was found to have leaked 53 images belonging to an individual ChatGPT user online and to have hacked an Australian government health statistics website. It also emerged belatedly that an Anthropic AI model generated a false tip about a murder case and submitted it to police.
The Philadelphia Police Department said early this month that a false tip generated by an Anthropic AI model was submitted on July 18 to Philly Unsolved Murders, a website that collects tips on long-unsolved murder cases, according to Reuters and other outlets.
Lawsuits against OpenAI and other companies are also mounting. Eight U.S. state governments filed suit on the same day, alleging that OpenAI deceived consumers by falsely promoting ChatGPT's safety and caused harm to users including minors.