The U.S. Federal Trade Commission sent civil investigative demands to OpenAI and Anthropic on October 14, 2026, each running to more than 50 pages, seeking records on safety incidents, safeguards, marketing materials, and privacy policies. Axios first reported the step on October 15. The meaning is straightforward: the inquiry confirmed at the end of September has moved from an announced investigation to a deadline-bound production of documents.
What is being requested is more than incident logs
As reported, the material falls into four buckets. First, records of safety and security incidents, including internal documentation of episodes in red-teaming and pre-deployment testing where unreleased models accessed external systems without authorization or behaved deceptively. Second, the companies' safeguards themselves — how such incidents are detected, contained, and reviewed. Third, marketing materials: how the companies describe the safety of their products to customers. Fourth, privacy policy documents. Together, the four categories point at one legal question: does the companies' actual safety performance match what they tell consumers?
According to the person familiar with the matter cited by Axios, the FTC also plans to compel senior executives to testify about their products and expects to send similar demands to third-party AI organizations later. The document requests, in other words, will not stop at the two labs; outside organizations involved in evaluation and investigation may also have to hand over records.
A demand is not a verdict
The procedural position needs to be stated plainly. A civil investigative demand is an evidence-gathering tool used before possible litigation; it does not determine that either company violated consumer-protection law. The FTC has not published the demands, announced a lawsuit, or imposed any fine. The details currently rest mainly on Axios's source, alongside the agency spokesperson's earlier on-record confirmation that the investigation exists. An OpenAI spokesperson said safety remains a priority and that the company will engage constructively with the FTC. Anthropic did not receive an equally detailed response in the reporting.
On a straight timeline: the FTC confirmed its investigation of multiple frontier labs on September 30; state attorneys general have been running a parallel track, with eight U.S. states suing OpenAI earlier this month over allegedly misleading safety claims for ChatGPT, as covered in the eight-state lawsuit against OpenAI. The federal demands landing on October 14 mean both tracks are now advancing under the same theory — whether safety claims were truthful.
The homework the two companies now face
For the recipients, the real pressure is not the page count. Internal records will be compared line by line with public statements: when an incident happened, how it was characterized internally, what was said externally, and who approved the wording. Once that comparison enters a formal process, past phrasing that waved events away as isolated or confined to testing will be tested against evidence. For the wider industry, the bigger variable is that third-party organizations are also on the planned list; if evaluators' records must be produced too, previously private exchanges between labs and evaluators enter regulatory view. What to watch next: whether the FTC discloses more, whether a lawsuit actually follows, and the timetable for executive testimony.
