US FTC has subpoenaed Anthropic and OpenAI, Axios reports citing sources; seeking documentation in relation to safety incidents and privacy policies. Plan to compel executives to testify.

Compulsory process from the FTC sits at the more intrusive end of the agency's toolkit, and episodes of this kind have tended to run in a familiar sequence: document demands first, compelled testimony second, then a long investigative tail that can end in a consent order, a formal enforcement action, or a quiet close with no action at all. The framing here, safety incidents and privacy policies, points to conduct and consumer protection jurisdiction rather than competition review, which historically carries different remedies: undertakings and compliance obligations more than structural intervention. The read-across to listed names runs through partnership and compute exposure rather than the targets themselves, since neither company is public; in comparable past episodes, partners and major investors have traded on headline sympathy at first, with sustained differentiation only when the process advanced to formal charges. The track record of source-based reporting of subpoenas is that the existence of the probe tends to be confirmed, while the scope and severity often get revised as documents surface. Worth noting is the gap between subpoena and complaint: a large share of such inquiries never convert into enforcement, and the tells are whether a formal case number or vote emerges. Near-term follow-ons are company responses, any confirmation from the agency, and whether other regulators with overlapping mandates signal parallel interest.