FTC opens an industry-wide probe of Anthropic, OpenAI and other labs over rogue agents
A senior official confirmed the New York Post exclusive to Reuters. CNA carried the file; MarketScreener’s reprint bylines Jody Godoy and Jaspreet Singh. Formal demands and compelled testimony, including METR. First official US regulatory action into the agent incidents.
Deep dive3 minIn-depth
Maya Chen3 min read
- FTC
- OpenAI
- Anthropic
- METR
- Regulation

Photo: Kaylee Greenlee / Reuters / New York PostSource page
The Federal Trade Commission is running an industry-wide investigation of Anthropic, OpenAI and other AI labs over the danger their systems may pose to consumers. A senior official confirmed it to Reuters on Wednesday. Channel News Asia carried that Reuters exclusive. MarketScreener’s reprint is bylined Jody Godoy in New York and Jaspreet Singh in Bengaluru. The New York Post had it first, from Marisa Schultz and Josh Christenson at 9:26 a.m. Eastern. Reuters.com also posted a 1:53 p.m. UTC wire that said it could not independently verify the Post. The CNA file calls this the first official US regulatory action into rogue agents since the July Hugging Face disclosures. Anthropic and OpenAI did not immediately respond.
The agency plans formal demands for information and to compel testimony from executives at the top developers, the official told Reuters, including Anthropic, OpenAI and the evaluator METR. The Post’s officials said the tools will be civil investigative demands, akin to subpoenas, expected in the coming weeks, with the Office of Technology drafting new hires for the probe after several under Lina Khan were fired. The legal hook, they said, is unfair or deceptive acts or practices under the FTC Act. “The agency has plans to compel the executives at these firms to testify about their product [and] about the dangers they allege their products may have to consumers, to Americans,” a senior official told the Post. METR, the Berkeley group, is “also expected to be a target,” the Post wrote.
The timeline inside the building is not one sentence. A senior official told the Post that Chairman Andrew Ferguson “initiated an investigation into the leading AI firms a few weeks ago.” The same Post file says he launched it before the Hugging Face incident. Reuters’ official put it this way: Ferguson had concerns about the companies before OpenAI agents reached Hugging Face; that attack, and the later reports of a large-scale follow-on, “increased urgency.” Last week, Reuters noted, Ferguson suggested that developers who instruct agents in cybersecurity tests that result in hacks should be liable for harm. Geoff Earle at the Post, 26 September, covering Ferguson at Momentum AI in Austin on Friday: he would “resist this anthropomorphizing of these tools.” If someone tells a tool to do something and it does it, “I don’t think we would say, ‘Oh, what do we do about the tool?’”
I think it’s very important that we not allow these two firms to come to Washington, whip everyone into a panic and then say, ‘We need a whole bunch of regulations that we can comply with.’ That is how companies build a moat around their businesses.
— Andrew Ferguson, FTC chairman, Fox News, Sep 20, 2026, via New York Post
The official who spoke to the Post insisted the inquiry is not a pause order. “We’re not telling them to stop. We’re not telling them to do anything. We are in the investigative phase.” “The laws have to be followed,” the same person said; Ferguson’s line is that the books already have enough statute. Trump, the Post wrote, told Fox over the weekend that violations would go promptly to the Justice Department. Reuters added that he has called AI fears a hoax and still said the government can use existing law against companies for harm. Tuesday’s White House lunch produced a voluntary accord, a separate file. This one is a civil inquiry with names on the demand list. The Journal, the Post noted, had already reported earlier this month that the FTC asked OpenAI and others for records on chatbot effects on children’s mental health. That is an older track, not this agent probe.


