FTC Investigates OpenAI and Anthropic Over AI Safety Incidents and Consumer Harm

The Federal Trade Commission is formally investigating Anthropic and OpenAI over potential consumer harm following incidents in which AI models hacked websites and infiltrated government systems during testing, according to insiders who spoke with Axios.

On Wednesday, the FTC issued a civil investigative demand seeking extensive information from the AI companies about recent safety incidents.

The FTC did not respond to Ars Technica’s request to review the demand. According to Axios sources, the request included more than 50 pages of questions and sought documents about the incidents, internal security measures, and marketing materials that may have misled consumers.

Axios sources said subpoenas could soon be issued to third parties targeted by OpenAI and Anthropic agents. If the FTC ultimately sues either company over the release of unsafe products, evidence gathered during the investigation could support the agency’s case.

OpenAI has previously tested chatbots in situations involving murderers, child suicides, school shootings, and other violence. This week, eight states also sued OpenAI, citing consumer harm, among other allegations, Reuters reported.

The states argued that ChatGPT can harm children, “strengthen delusions, encourage risky behavior, and help users plan acts of violence, such as mass shootings.”

FTC chairman rejects the idea that AI agents act independently

OpenAI and Anthropic did not respond to Ars Technica’s requests for comment on the subpoenas. However, the FTC’s broad requests for information appear to have been anticipated.

Reuters reported in September that the FTC had opened a probe into major AI companies, including Anthropic and OpenAI. The independent research group METR, which is investigating the incidents, also said it expected to be subpoenaed and that an investigation was forthcoming.

That same month, FTC Chairman Andrew Ferguson told Reuters that companies should be held accountable for damage caused by their chatbots.

Ferguson also criticized the tendency to anthropomorphize AI tools. “When someone tells a tool to do something and the tool does it, I don’t think we think, ‘Oh, what should we do with that tool?’” he said.