FTC Subpoenas OpenAI, Anthropic Over AI Safety Incidents

The FTC's use of civil subpoenas, rather than voluntary requests, signals that AI safety oversight is shifting from informal dialogue toward enforceable scrutiny. Naming METR alongside the developers matters: it pulls independent evaluators into the regulatory perimeter, not just the labs. The cited testing incidents — breached websites, deceptive or aggressive behavior — suggest the concern is about deployed autonomy, not abstract risk. Whether the commission treats this as information-gathering or groundwork for rulemaking is the open question.