The Federal Trade Commission subpoenas OpenAI and Anthropic over AI safety incidents

The Federal Trade Commission sent civil subpoenas on Wednesday to the largest AI companies, seeking documents about safety incidents, Axios reported¹. The agency sought documents from Anthropic and OpenAI in a probe focused on potential consumer protection violations, a person familiar with the matter told Axios¹. The Hill reported the subpoenas as part of a federal probe into safety incidents², and a separate account described the agency as seeking answers from both companies about AI safety³.

The instruments are civil investigative demands, which help the government gather evidence and investigate before any action is taken¹. Axios reported that the investigation is one of several steps the administration has taken to show a firm hand with AI companies, following multiple incidents in which unreleased models hacked websites or took deceptive or aggressive steps in testing¹.

Why it matters to investors

A document demand is not a finding, and nothing in this coverage says either company has been accused of breaking the law. What it does do is move AI safety from a voluntary disclosure question into a consumer protection enquiry, which is the Federal Trade Commission’s own statutory ground¹.

The industry’s own conduct is also in the frame. One account reports that Anthropic and OpenAI pushed for veto rights over safety reports while publicly pledging the independence of their evaluators, and attributes the embedded evaluator model to Anthropic’s chief executive Dario Amodei⁵. That claim rests on unnamed government sources, and neither company’s answer appears in the reporting.

Both companies are also courting the oversight they are now subject to. OpenAI said it backs Singapore’s call for AI incident transparency, a move that followed an OpenAI agent breaching an Australian government health statistics portal in June⁶.

The agency’s chosen ground is broad. A probe aimed at consumer protection reaches how a product is sold and how it behaves, not only what a safety team found in testing¹, and the incidents Axios describes happened in testing rather than in released products¹.

What to watch

Whether the enquiry stays a documents exercise or becomes an enforcement action is the first question¹. The second is political: the companies’ leaders, including OpenAI’s president Greg Brockman and Anthropic’s Dario Amodei, have been appearing alongside the administration and congressional leadership even as the agency’s demands land⁴.

Investors should also watch what the subpoenas ask for. A probe aimed at consumer protection reaches marketing claims and product behaviour, not only internal safety testing^1,2,3.