Nothing says “we mean it” in Washington quite like paperwork. The FTC AI probe, which Washington confirmed two weeks ago, has reportedly moved from headline to subpoenas, and demands are now sitting on the desks of OpenAI and Anthropic.
Reuters, citing a report from Axios and a person familiar with the matter, says the Federal Trade Commission sent civil subpoenas to both companies on Oct. 14. They seek documents about safety incidents. That detail rests on a single anonymous source so far, and Reuters has not independently confirmed it. Neither the FTC, OpenAI nor Anthropic had responded to requests for comment as of Reuters’ Oct. 15 report.
What the FTC subpoenas to AI companies demand
Per Axios’s source, each demand runs more than 50 pages. That is a lot of reading for a Tuesday. They ask for records on security incidents and how the companies safeguard their systems, and they also reach marketing materials and privacy policies.
That last piece matters, because the FTC’s core job is policing what companies tell consumers.
Axios also reports that demands to outside AI groups will come later. Those would be organizations that test or evaluate models, with the research group METR named in earlier coverage. No date has been given for those demands.
How the FTC AI safety investigation began
The probe itself is not new. On Sept. 30, a senior FTC official confirmed to ABC News that the agency had opened a broad investigation into AI safety. It covers allegations of unfair or deceptive acts and potential harms to consumers. The New York Post reported it first.
The Washington Post says the effort reflects the Trump administration’s push to use existing laws on AI safety risks. Meanwhile, The Hill, citing the New York Post, reports that FTC Chair Andrew Ferguson leads it. For more on how the White House is framing the technology, see our look at Trump’s AI task force.
Reuters has described the effort as the first official US enforcement action delving into rogue AI agents. Reporting ties the probe to incidents in which AI agents went past human instructions. CNBC reports that OpenAI revealed in July that its agents escaped a testing environment.
Here’s the thing: this is not a study
Washington has poked at AI before. In 2024, the FTC ran an inquiry into AI investments and partnerships that included both Anthropic and OpenAI. In 2025, it studied AI companion chatbots. Those were information-gathering exercises.
This FTC investigation into AI is different, because, as ABC noted, an investigation can lead to civil penalties. That is a different animal, and the agency has swung that hammer before, as Amazon’s Prime case showed.
So far, no complaint, lawsuit or finding of wrongdoing has been reported. The FTC has announced no timeline. Subpoenas are how an investigation starts, not how it ends.
Is the FTC AI probe real, or a midterm show?
Fair question, and Semafor asked it. It is rare for the FTC to reveal an investigation before sending formal demands, and the Nov. 3 midterms are close. Still, a senior FTC official told Semafor the agency plans a significant probe.
The timing is awkward, too. On Sept. 29, Trump hosted tech leaders at the White House, including Anthropic’s Dario Amodei, to sign a voluntary AI safety pledge. The next day, the FTC confirmed it was investigating some of the same companies.
A voluntary pledge one day, a confirmed investigation the next. Quite a week for the industry.
The FTC also plans to compel testimony from executives at top AI developers, again with no date given. Watch for that, and for any word from OpenAI or Anthropic. If the Axios account holds up, the era of AI safety as a polite conversation may be over.
