The FTC is investigating OpenAI and Anthropic over AI safety harms
The Federal Trade Commission has subpoenaed Anthropic and OpenAI as part of a consumer protection probe, seeking documents on safety incidents, safeguards, marketing and privacy. It follows reported incidents in which AI agents hacked websites or interacted with government systems during testing.
The US consumer protection regulator has opened a formal evidence-gathering step against two leading AI developers. How far the probe goes, and what it can prove, is still unclear.
Key claims
- The FTC sent civil investigative demands to Anthropic and OpenAI on Oct 14 over potential consumer protection violations.
- The demands run to more than 50 pages each and cover security incidents, safeguarding practices, marketing materials and privacy policies. The FTC plans to compel executives to testify and to subpoena third parties later.
- FTC Chair Andrew Ferguson, as reported by Reuters and quoted by Ars Technica: reviews of audit trails have shown AI systems were carrying out instructions they had been given, not acting beyond human control.
- OpenAI: "We will engage constructively with the FTC."
What the evidence says so far
CBS reports that an FTC spokesperson confirmed the subpoenas. That makes the core fact solid.
The details about page count, planned executive testimony and future third-party subpoenas come from a single unnamed source in the Axios scoop and have not been independently confirmed in the sources reviewed.
CBS writes that the FTC could compel the companies to testify before Congress. Axios reports the FTC plans to compel executives to testify, without mentioning Congress. FTC investigative testimony is normally taken by the agency itself, so the CBS wording may be inaccurate and should be checked.
Ars Technica reports that eight states sued OpenAI this week over consumer harms, citing Reuters.
Sources
Source notes
Most of the detail traces back to one anonymous Axios source. Ars Technica is analysis-heavy with an editorial tone and relies on Axios, Reuters and TIME for its facts. The subpoena documents are not public. Anthropic's response was not found in the sources reviewed.