FTC reportedly opens AI product-safety probe into OpenAI and Anthropic

Two news reports say the Federal Trade Commission has opened a broad investigation into the safety of artificial intelligence systems made by OpenAI and Anthropic, citing people familiar with the confidential inquiry. The reports do not establish that the agency has brought an enforcement action or found that either company broke the law.

The inquiry could bring the companies’ product-safety claims and practices under scrutiny through the FTC’s existing consumer-protection authority. One report said the agency was preparing information demands likely to be sent in the coming weeks. The available reporting did not establish that any demands had been issued or served, nor did it identify the other companies reportedly under scrutiny.

By the research cutoff, no FTC complaint, consent order, administrative docket, resolution, civil investigative demand, closing letter or press release tied to the reported September 2026 probe had been located. The products, conduct, incidents, legal theories and time period under examination remain undisclosed. FTC investigations can close without public action.

Formal information demands are investigative tools, not charges. Under the FTC’s investigative process, the agency can seek documents, written answers, testimony and other information relevant to potential unfair or deceptive practices before beginning a proceeding. Receiving a demand does not itself mean the recipient violated the law.

Section 5 of the FTC Act prohibits unfair or deceptive acts or practices in commerce. The agency’s unfairness policy says a practice must cause or be likely to cause substantial consumer injury that people could not reasonably avoid and that is not outweighed by benefits to consumers or competition. Unwarranted health and safety risks can qualify; trivial or speculative harm does not.

The agency’s deception framework examines whether a representation, omission or practice is likely to mislead a reasonable consumer and whether it is material to a decision about a product or service. That framework could apply to materially misleading AI safety claims or omissions. Design or deployment practices could fall under unfairness only if the consumer-injury test is met. No public record reviewed for this story identifies either theory as the basis for the reported investigation.

A proposed FTC policy statement issued in July 2026 says AI products and services remain subject to Section 5. It distinguishes model mistakes from deceptive conduct: hallucinations alone do not raise an issue under laws the FTC enforces, the statement says, while misrepresenting how likely they are may constitute deception.

The reported probe is separate from the FTC’s September 2025 study of consumer-facing AI companion chatbots. That public inquiry included OpenAI but not Anthropic and used the agency’s Section 6(b) study authority, which the FTC said had no specific law-enforcement purpose.

The investigation is also distinct from voluntary industry coordination. OpenAI’s policy chief said in September that the company was in AI safety talks with Anthropic and Google DeepMind, but no agreement or shared standards had been disclosed.

According to the two reports, OpenAI had no immediate comment on the product-safety inquiry, while Anthropic did not immediately respond to requests for comment.