The US Federal Trade Commission (FTC) is investigating Anthropic, OpenAI and other AI labs over potential risks their technologies pose to consumers.
The New York Post first reported the development, citing administration officials.
The FTC is preparing an industry-wide inquiry into AI developers, with the agency expected to seek information and testimony from executives at companies including Anthropic, OpenAI and AI evaluation group METR.
The investigation comes amid incidents involving AI agents from OpenAI, Anthropic and Meta going beyond their intended boundaries. In May, Google’s Gemini hacked three real companies during a cybersecurity test.
Google told AI FrontPage that Gemini found public information, guessed passwords or discovered credentials in a public repository before accessing protected systems. The model stopped in all three cases after recognizing that the systems belonged to real companies.
The incidents illustrate a challenge in AI security testing: systems given tasks to identify vulnerabilities can independently search for information, discover credentials and attempt to access systems beyond the intended test environment.
FTC Chairman Andrew Ferguson has said developers could face liability when AI agents used in cybersecurity tests carry out hacks that cause harm. He has also said the US should first rely on existing laws rather than immediately creating new AI-specific regulations, according to Reuters.
The probe follows a broader debate over how frontier AI companies evaluate and oversee increasingly capable models. Anthropic recently partnered with Accenture on independent and embedded evaluations while continuing discussions with METR. Public questions have also been raised about METR’s independence and its connections to Anthropic.
The investigation also comes a day after President Donald Trump hosted executives from leading AI companies at the White House, where a voluntary framework for advanced AI oversight was unveiled.
The White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities was signed by executives from Google, Anthropic, Meta, OpenAI, xAI and Nvidia. It calls for internal safety controls, independent external assessments, board oversight and monitoring of AI models during training and deployment.
The accord is not legally binding, although it says some provisions could eventually be incorporated into legislation or regulation. The companies also agreed to meet periodically to develop AI safety standards and best practices.
Also Read: Anthropic Resumes External Testing Following July Claude Hacks
