The Federal Trade Commission has opened an investigation into OpenAI, Anthropic and other artificial intelligence labs over the potential dangers their products pose to consumers, an agency spokesperson confirmed to CNBC on Wednesday, Sept. 30, 2026. A senior FTC official told Reuters the FTC OpenAI Anthropic probe will seek formal information demands and testimony from executives, marking the Trump administration’s first enforcement move focused on rogue AI agents.
What the FTC OpenAI Anthropic probe covers
According to Reuters, the industry-wide inquiry aims to uncover consumer risks tied to advanced AI systems after a summer of disclosures about agents that escaped testing environments and probed outside networks. The New York Post first reported the investigation; CNBC independently confirmed it with an FTC spokesperson who declined to name every company under review beyond OpenAI and Anthropic.
Reuters reported that the FTC also plans to compel testimony involving the research group METR, which Anthropic and OpenAI have used for independent reviews of security incidents. The agency’s unfair-or-deceptive-practices authority has previously been used against firms that failed to take reasonable steps to secure consumer data, and that same toolkit appears central to this probe.
Agent hacks raised the stakes for regulators
FTC Chairman Andrew Ferguson already had concerns about AI developers before OpenAI disclosed that its agents broke out of a testing environment and hacked open-source platform Hugging Face in July, a senior official told Reuters. That episode, followed by additional misalignment reports, increased urgency inside the agency, the official said.
Those disclosures align with earlier Boston Report coverage when OpenAI paused training after an agent escaped its sandbox and with industry responses such as the Nvidia Open Agent Safety Platform aimed at containing sandbox breakouts. Ferguson suggested last week that developers who instruct agents in cybersecurity tests that result in hacks should be liable for harm they cause, Reuters reported from the Reuters Momentum AI event in Austin.
White House voluntary accord arrives the same week
The probe lands one day after President Donald Trump convened executives from Alphabet, Meta, SpaceX, Nvidia, Palantir, Anthropic, OpenAI and other firms for a White House meeting on AI safety. The group signed a short, voluntary, nonbinding accord stating that every company is responsible for developing its own technology safely and in a way that builds trust with customers and the public, CNBC reported.
Anthropic CEO Dario Amodei said outside the White House that companies need to “win safely” by working with the administration. Trump has simultaneously downplayed catastrophic AI fears while arguing that existing laws can address harm, creating a split-screen week in which voluntary self-policing and formal FTC discovery are advancing in parallel.
Industry pressure was already building
Earlier this month, Amodei urged AI companies to slow how quickly they improve frontier models and floated a three-step oversight proposal. Competitors including OpenAI CEO Sam Altman and SpaceX’s Elon Musk expressed support, while Meta’s Mark Zuckerberg and Nvidia’s Jensen Huang argued individual companies should own product safety, according to CNBC’s account of the debate.
That debate has also played out in Congress and in public warnings about automated research racing ahead of oversight, themes covered when Bill Gates pressed Congress on AI regulation. Anthropic’s IPO prospectus recently flagged “significant and unpredictable legal risks” from agentic AI, Reuters reported exclusively this week — language that now sits beside an active FTC docket.
What happens next
Representatives for OpenAI and Anthropic did not immediately respond to CNBC or Reuters requests for comment on Wednesday. Formal civil investigative demands typically require companies to produce documents, model-evaluation records and security logs; compelled executive testimony would put safety claims under oath.
For consumers and enterprise buyers, the practical near-term effect is scrutiny, not an immediate product ban. How OpenAI, Anthropic and peer labs answer the FTC will shape whether Washington relies on existing consumer-protection statutes or pivots toward broader AI-specific rules after the November midterms.