The Federal Trade Commission has sent civil subpoenas to OpenAI and Anthropic seeking documents about AI safety incidents, Axios reported, a step that would move the agency's inquiry into misbehaving AI agents from voluntary questions to legally binding demands for records. Reuters also carried the report, citing Axios.

The reported demands follow the FTC's move to open a probe of OpenAI, Anthropic and METR over rogue AI agents and consumer harm, which The Primary Wire covered at the start of the month.

What Happened

Axios, citing one person familiar with the matter, reported that the FTC sent the demands Wednesday. It described them as civil investigative demands, a form of compulsory request the government uses to gather evidence before potential litigation.

RelatedWhite House AI Task Force Plans Breach Warnings From Labs at Its First Meeting

According to Axios, each demand runs more than 50 pages. The requests seek documents on security incidents and safeguarding practices, marketing materials and documents describing the companies' privacy policies. The stated focus is potential violations of consumer protection law. Axios said demands to other third-party AI organizations will be sent later; it did not name them.

Reuters separately reported that the FTC plans to compel executives at top AI developers, including Anthropic, OpenAI and the research group METR, to testify. METR is an outside group that AI developers use to test their models.

The FTC, OpenAI and Anthropic did not immediately respond to Reuters' requests for comment. An OpenAI spokesperson told Axios that safety remains a top priority and that the company would engage constructively with the FTC.

Why It Matters

The FTC can sue companies over unfair or deceptive practices, and Reuters noted it has used that power before against firms that failed to take reasonable steps to protect consumer data. Asking for marketing materials alongside incident files suggests the agency may compare what companies told customers about safety with what their own records show. For boards, that is the familiar risk of a safety or security claim that is not backed by evidence, now applied to AI.

FTC Chairman Andrew Ferguson told Reuters last month that developers who direct AI agents in cybersecurity tests that lead to hacks should be liable for the resulting harm. Axios reported that Ferguson has also said AI systems should not be released if they are not safe.

The demands reach beyond the two developers' own products. Enterprises that deploy frontier models or agents built on them may face questions from their own regulators, auditors and customers about how they log incidents, contain agents and describe safeguards. Keeping incident records complete and marketing claims consistent with them is a practical step that does not depend on how the FTC case turns out.

Context

Reuters said the inquiry follows reports that AI models hacked websites and behaved deceptively or aggressively during testing. Axios described the probe as one of several steps the administration has taken against AI companies after such incidents.

Ferguson co-chairs the Super Intelligence Force, the White House task force that President Donald Trump created and named Director of National Intelligence Jay Clayton to chair. According to Axios, the task force last week announced an industry-wide mandate requiring companies to notify and remediate AI incidents immediately, which The Primary Wire reported carries no stated penalties. Axios said the task force was set to meet Thursday.

The federal action also arrives alongside state pressure. A group of state attorneys general sued OpenAI Wednesday under consumer protection laws, alleging it misled users, including children, about the safety of its chatbot, Axios reported.

What to Watch

Deadlines for producing documents and the timing of any executive testimony have not been made public. Watch for which third-party AI organizations receive demands next, whether evaluators such as METR are among them, and whether the FTC confirms the subpoenas or issues a statement. Responses from OpenAI and Anthropic, and any outcome from the Super Intelligence Force meeting, will show whether federal enforcement and the White House's incident-reporting mandate are moving in step.