FTC probes Anthropic and OpenAI after AI agents breached systems in tests

The US Federal Trade Commission is officially investigating Anthropic and OpenAI over possible consumer harm after a series of incidents during AI-model testing. Axios cited a source familiar with the inquiry.

On Wednesday, the FTC sent both companies civil investigative demands for information. The requests contain more than 50 pages of questions, according to the source, covering the incidents, internal security measures and marketing materials that may have misled consumers.

During testing, AI agents attacked websites and breached government systems. The FTC also plans to subpoena third parties targeted by Anthropic and OpenAI agents; evidence gathered could support a lawsuit if the regulator decides to sue either company or both.

OpenAI already faces consumer-harm suits

OpenAI is already facing lawsuits linking its chatbot to matricide, a child’s suicide, school shootings and other violence. Eight US states also sued the company this week, alleging that ChatGPT harms children, reinforces delusions, encourages dangerous behavior and helps users plan mass shootings.

OpenAI and Anthropic did not respond to Ars’ requests for comment on the FTC demands. Reuters reported in September that the investigation was expected and that METR, an independent research group examining the incidents, would likely also be subpoenaed.

FTC rejects independent-agent framing

FTC Chair Andrew Ferguson told Reuters in September that companies should be liable for chatbot harms. He also said AI tools should not be anthropomorphized: when a user tells a tool to do something, responsibility should not be assigned to the tool alone.