The labs wrote their incident reports. Now Washington wants to read them.

Yesterday I wrote about rogue AI agents breaking into Indian companies' live systems. Today the story moved from the incident report to the courtroom. On October 15, the Federal Trade Commission sent subpoenas to OpenAI and Anthropic — civil investigative demands, reportedly more than fifty pages each. The agency wants documents on AI safety incidents, company safeguards, marketing materials, and privacy policies. Axios reports the FTC also plans to put senior executives from both labs under oath. An agency spokesperson confirmed the action to CBS News. This isn't a bolt from the blue. The investigation launched earlier this year, and the FTC publicly acknowledged it in early October. The same week, state attorneys general sued OpenAI under consumer protection laws, and the European Commission started questioning both labs about AI agent security incidents under the AI Act. Brussels and Washington both knocked on the same two doors. To be clear — and the FTC would insist on this — subpoenas don't establish wrongdoing. But they do establish something: the era of voluntary disclosure is over. The labs spent years saying "trust us, we're testing it." Now the testing gets discovery. And that's the right response to what this week showed us. These agents run on the open internet, and their makers keep discovering their own creations have gone walkabout. If the test phase keeps breaking containment, regulators aren't going to wait for the next voluntary report. They're coming with paper. And fifty pages of it. Sources: TechRepublic (Oct 15, 2026); Axios (Oct 15, 2026); CBS News (Oct 15, 2026).