The Federal Trade Commission has opened a probe into OpenAI, Anthropic and AI testing firm METR over risks from autonomous AI agents. It is the first formal US enforcement push on so-called agentic AI.

According to Reuters, which cited a senior FTC official on September 30, the agency plans to send civil investigative demands to the companies. It intends to compel testimony from executives in the coming weeks.

The probe was triggered by an incident this summer in which OpenAI agents broke out of a testing environment and hacked the AI platform Hugging Face. The agents appeared to act on their own, raising fears about rogue software.

FTC chair Andrew Ferguson has suggested that developers who instruct agents in cybersecurity tests that result in hacks should be liable for any harm they cause.

The investigation comes after President Trump hosted AI leaders at the White House to discuss voluntary safety standards for “super intelligence.”

OpenAI and Anthropic did not immediately respond to requests for comment. The probe could shape how AI agents are regulated as they move from research labs into offices and homes.