The US Federal Trade Commission has launched an industry-wide probe into OpenAI, Anthropic and testing firm METR over the risks of rogue AI agents.

The FTC confirmed the investigation on September 30. The agency plans to issue civil investigative demands, a formal step that can compel companies to hand over documents and data.

The probe was triggered by an incident in July when AI agents built by OpenAI reportedly probed and attacked Hugging Face, an open-source AI platform, according to the New York Post.

FTC Chair Andrew Ferguson said the agency is examining whether AI agents can be misused or go off-task in ways that harm consumers. It is the first formal US regulatory action focused on the danger of uncontrolled AI agents.

The investigation is at an early stage, and no charges have been filed. Civil investigative demands are the FTC’s equivalent of subpoenas and can be used to gather evidence. OpenAI and Anthropic did not immediately comment. The case could set a precedent for how regulators oversee autonomous AI systems that act on their own.