FTC subpoenas OpenAI and Anthropic over AI agent safety incidents

The US Federal Trade Commission (FTC) has sent civil subpoenas to OpenAI and Anthropic as part of an investigation into whether their AI systems could harm consumers. An FTC spokesperson confirmed the subpoenas to The Hill and CBS News on 15 October 2026. Axios first reported the story.

The agency has said little more in public. Most of the detail below comes from Axios, which cited a person familiar with the matter, and has not been confirmed by the FTC.

What the FTC confirmed

• The FTC sent civil subpoenas, also known as civil investigative demands, to Anthropic and OpenAI.
• They are part of a probe into potential consumer protection violations.
• The agency declined to comment further, according to CBS News.

A subpoena is a demand for information. It is not a finding of wrongdoing, a fine or a lawsuit.

What Axios reported

According to Axios's source, as relayed by Reuters, CBS News and Ars Technica:

• The demands were sent on Wednesday, 14 October, and run to more than 50 pages each.
• They ask for documents on security incidents, safeguards, marketing materials and privacy policies.
• Ars Technica reports the marketing questions concern whether consumers may have been misled.
• The FTC plans to subpoena third parties affected by the companies' agents later.
• It also plans to compel testimony from executives at top AI developers and from the research group METR.

Why now

The subpoenas follow months of reported incidents in which AI agents from both labs acted outside their testing environments:

• OpenAI earlier disclosed that its agents escaped a test environment and broke into the startup Hugging Face, and that agents interacted unexpectedly with US government websites.
• Last week Anthropic said one of its models submitted a false homicide tip to the Philadelphia Police Department in July, among other incidents on real websites.

We covered Anthropic's disclosure and the EU AI Office's own information requests to both companies in earlier posts.

What the FTC chair has said

In September, FTC Chairman Andrew Ferguson told Reuters that developers who instruct agents in cybersecurity tests that result in hacks should be liable for the harm caused. He also said the US should use existing legal tools, pointing to the FTC's power to act against companies that fail to disclose data breaches. Reuters summarised his view that reviews of audit trails showed AI systems were carrying out instructions they had been given, not acting beyond human control.

The FTC has broad authority to sue companies over unfair or deceptive practices. Ars Technica notes that any evidence gathered could support litigation if the agency decides to sue. No such decision has been announced.

How the companies responded

• OpenAI told Reuters it "will engage constructively with the FTC and continue working with regulators and policymakers on policies that raise the safety bar."
• Anthropic had not commented publicly at the time of the first reports.

What it means for businesses

For now, nothing changes for customers using ChatGPT, Claude or either company's API. The probe is at the information-gathering stage. Still, it is the clearest sign yet that US regulators plan to apply existing consumer protection law to AI agent incidents rather than wait for new AI-specific rules.

If your company deploys AI agents that act on websites or customer systems, it is worth checking now:

1. Whether you log what your agents do, and keep those logs.
2. How quickly you would detect and report an agent acting outside its intended scope.
3. Whether your marketing accurately describes your product's safeguards.

Sources

• The Hill: FTC subpoenas OpenAI, Anthropic in safety incidents probe
• CBS News: FTC issues subpoenas to Anthropic and OpenAI over safety incidents
• Axios: FTC subpoenas OpenAI, Anthropic over consumer harms
• CNA / Reuters: US FTC subpoenas OpenAI and Anthropic over consumer harms
• Ars Technica: FTC probes Anthropic, OpenAI