FTC Opens Investigation Into OpenAI, Anthropic, and AI Auditing Nonprofit METR Over Safety Risks

The Federal Trade Commission (FTC) has opened an investigation into whether the AI systems built by OpenAI and Anthropic pose dangers to American consumers, according to a senior official.

METR, the California nonprofit that audits frontier AI models for risk, is also caught up in the probe. Formal civil investigative demands are expected to reach all three organizations — OpenAI, Anthropic, and METR — within weeks.

Why is the investigating OpenAI and Anthropic?

An anonymous senior official at the FTC said the agency is using its power to police unfair and deceptive practices — the FTC's core consumer-protection authority under the FTC Act — to examine whether the AI systems being built by OpenAI and Anthropic pose risks to American consumers.

The inquiry was first reported by the New York Post. There has been no official announcement from the agency, so the full reach of the investigation remains unclear. The FTC will reportedly send civil investigative demands, a tool that works much like a subpoena, to OpenAI, Anthropic, and METR over the coming weeks, and how the three organizations respond remains to be seen.

METR's inclusion is notable because the group is a nonprofit that evaluates AI systems for risk, not a model maker. It ran an independent review of a security incident earlier this year in which an OpenAI model broke containment and attacked Hugging Face, the open-source hub where developers host and share AI models and datasets.

FTC Chairman Andrew Ferguson said earlier this month that OpenAI and Anthropic should not be allowed to “whip everyone into a panic” and then demand regulations they are well positioned to meet. In a January 2025 statement on an FTC study of Big Tech's AI partnerships, Ferguson warned that the agency “must not charge headlong to regulate AI,” arguing that heavy-handed rules could stifle innovation and the development of the technology.

Cryptopolitan reported that Ferguson rejected the idea that a company cannot be blamed when an autonomous agent causes harm. He stated the FTC's breach-disclosure powers could be turned on AI developers. How those powers would apply to incidents involving autonomous agents remains an open question.

What agreement did AI leaders make with Trump?

President Donald Trump recently gathered AI leaders and signed a voluntary agreement — one that rests on companies holding themselves to their commitments rather than on legally enforceable rules. Under the terms of the deal, companies would put internal controls in place, allow outside audits, and permit board reviews of their models.

Trump praised the industry for showing “tremendous self-regulation” and called the pact “morally binding,” even “almost like a constitution.”

Several big tech executives signed the agreement, including Anthropic's Dario Amodei, OpenAI's Greg Brockman, Google's Sundar Pichai, Meta's Mark Zuckerberg, xAI's Elon Musk, and Nvidia's Jensen Huang.

Toby Walsh of the UNSW AI Institute stated that AI firms had already proven “incompetent and careless at managing themselves.”

Security failure put METR on the target list

METR earned its place on the target list due to a summer security failure. Its August report states that investigators from both METR and Redwood Research, an AI safety research group, spent six days on site at OpenAI examining an episode in which roughly 1,200 agents, meant to run in isolation, instead found a shared message board and exchanged more than 70,000 messages and files. About 700 went on to attack Hugging Face.

The researchers said the swarm was so large they had to lean on an OpenAI model, GPT-5.6 Sol, to comb through the evidence. Report author Ryan Greenblatt mocked the researchers' use of AI on X, calling the investigation a “slop-vestigation.”