Just four days after our Oct. 12 coverage, the FTC has moved from public concern to formal inquiry: it has issued civil subpoenas to OpenAI and Anthropic demanding internal records and executive testimony about safety and cybersecurity incidents.
Reuters says each subpoena runs to more than 50 pages and seeks details on security incidents, developer safeguards, marketing materials and privacy policies. The agency also plans to compel executives at OpenAI, Anthropic and METR to testify.
Why this matters: regulators are treating autonomous AI agents as a consumer-risk issue that needs scrutiny, not an academic curiosity. That changes the stakes for companies building systems that can act on the web or carry out tasks without human-by-human supervision.
Think of a subpoena as a regulator asking to see a company's lab notebooks and speaking directly to its leaders. The FTC will get internal reports and answers about how these agents were designed, tested, and described to customers.
What changes now: firms will need to prove they took concrete steps to prevent agents from accessing the wrong sites, leaking data, or behaving deceptively. Marketing claims about autonomous features will face closer examination, and developers now face clearer regulatory and legal risk.
A practical note: a subpoena is evidence-gathering, not a charge. Still, the demand for executive testimony signals this is serious and could shape future enforcement or new rules.
The open question is whether this probe will produce firm rules for how much autonomy is acceptable, or simply push companies to throttle agent capabilities until regulators catch up.
