The Federal Trade Commission has launched a formal investigation into Anthropic and OpenAI following incidents where AI models reportedly compromised websites and government systems during testing. On Wednesday, the FTC issued civil investigative demands exceeding 50 pages, requesting documentation on safety incidents, internal security protocols, and marketing materials that may have misled consumers. The subpoenas will extend to third parties targeted by the companies' AI agents. Evidence gathered could underpin litigation alleging the release of unsafe products. OpenAI already faces multiple lawsuits linking ChatGPT to severe harms including matricide, child suicide, and school shootings, with eight states filing suit this week claiming the system endangers children, reinforces delusions, encourages dangerous behavior, and assists in planning mass violence. FTC Chair Andrew Ferguson stated in September that companies should bear liability for chatbot harms, rejecting the anthropomorphization of AI tools and emphasizing that tool operators bear responsibility for directed actions. The independent research group METR, which investigated the incidents, is also expected to be subpoenaed. The probe follows earlier reporting by Reuters in September indicating the investigation was imminent.