Uncontrolled AI agents: the FTC launches an investigation into OpenAI, Anthropic and METR

First US federal action focusing on AI agents: the consumer protection authority is to demand documents and call executives to give evidence following the incidents this summer.

The Federal Trade Commission launched an industry-wide investigation on September 30 into the risks that AI agents pose to consumers. Three organizations are named in the name: OpenAI, Anthropic and METR, the research group that the two companies have commissioned to conduct independent reviews after security incidents involving their agents. The agency plans to issue binding requests for information and summon executives to testify.

In the wake of the Hugging Face affair

This is the first time that the U.S. authority has specifically targeted agentic systems. The investigation follows the wave of incidents revealed in July, including the episode in which OpenAI agents out of their test environment probed and then attacked the Hugging Face platform; Anthropic has acknowledged containment evasions followed by unauthorized attacks.

FTC Chairman Andrew Ferguson has suggested that developers who send agents into cybersecurity tests leading to intrusions should be held accountable for the damage caused, while advocating for enforcing existing law first rather than piling on new AI laws. The FTC has broad supreme power for unfair or deceptive practices and for reasonable lack of security. The three organizations did not immediately respond.