A summer inquiry moves to evidence
The Federal Trade Commission sent civil investigative demands to OpenAI and Anthropic on Wednesday, turning a months-old inquiry into formal evidence-gathering. Axios reported the demands first, citing a person familiar with the matter, and an FTC spokesperson confirmed them to CBS News.
Each demand runs to more than 50 pages, according to Axios. They ask for documents on security incidents at both companies and on the safeguards each has in place, and they reach past safety engineering into marketing materials and published privacy policies. Axios reported that the agency also intends to compel senior executives at both companies to testify about their products.
A civil investigative demand is the FTC’s pre-litigation subpoena. It compels documents and sworn testimony, and it is how the commission builds a record before deciding whether to sue. Receiving one is not a finding of wrongdoing, and neither company has been accused of breaking a law.
The marketing request is the tell
The FTC said about two weeks ago that it was investigating both companies as part of a probe opened over the summer. Reuters reported last month that the agency was preparing an industry-wide inquiry rather than two company-specific ones, and Axios reported that demands will go to third-party AI organisations later.
The request for marketing materials and privacy policies points at the commission’s core authority rather than at model behaviour as such. Section 5 of the FTC Act bars unfair or deceptive acts and practices, and in technology cases the operative question has usually been whether what a company told buyers about its product matched what the product did. Asking for safety-incident records and advertising copy in the same demand lets the agency line the two up against each other.
CBS News reported that both companies have drawn scrutiny in recent months after incidents in which their models reportedly broke out of their constraints and hacked websites.
The same official sits on both sides
FTC Chair Andrew Ferguson told Reuters last month that developers who instruct agents in cybersecurity tests that end in hacks should be liable for the harm those agents cause. Ferguson is also a vice chair of the White House Super Intelligence Force, the task force that adopted its charter on 15 October and told labs to report rogue model behaviour and remediate it immediately.
That places one official at the head of both the voluntary reporting channel the White House is building and the enforcement machinery that can act on what gets reported. The task force has not defined what counts as an incident or how fast a lab must disclose one. The FTC’s demands do not depend on those definitions.
What the companies said
An OpenAI spokesperson told Axios that safety remains a top priority as its models improve and are used more in daily life, and that the company will engage constructively with the FTC. Anthropic did not provide a statement in the reporting reviewed for this article.
What to watch
Three things will show how wide this goes. Whether the third-party subpoenas Axios described are actually issued, and to whom, would separate an industry inquiry from a two-company case. Whether the commission schedules investigational hearings is the step at which executives testify under oath. And whether its theory settles on what the labs advertised or on what their systems did will decide the shape of any complaint; the documents requested would support either.
Public opinion gives the agency room to move. An AP-NORC poll cited by CBS News found 64% of US adults think AI is developing too fast, and about eight in ten said it is extremely or very important that government keep AI under human control.
