FTC Opens Consumer-Protection Probe Into OpenAI, Anthropic, Forcing Document Handovers

The Federal Trade Commission has opened a formal investigation into OpenAI, Anthropic and other major artificial intelligence developers, examining whether their products violate consumer protection law. The agency is pursuing the inquiry through compulsory process, meaning it can force companies to turn over internal documents and put executives under oath. The probe predates the recent breach at Hugging Face, according to reporting from The Decoder, and signals that regulators are treating consumer harms in AI as a distinct enforcement priority rather than a side effect of existing tech rules.

The scope is broader than the FTC's earlier posture toward AI companies. That prior effort focused on holding developers accountable when their autonomous agents cause harm — a liability question tied to what a model does in the wild. This investigation instead targets the consumer-facing side of the business: how these labs market their systems, what they promise about accuracy and safety, and whether those claims hold up. Compulsory demands give the commission leverage that voluntary requests for information do not, because noncompliance can lead to court enforcement.

OpenAI and Anthropic are the two most prominent names attached to the inquiry, and both sit at the center of the consumer AI boom. OpenAI's ChatGPT and Anthropic's Claude are used by hundreds of millions of people, many of them paying subscribers who rely on vendor claims about reliability and data handling. The FTC has not publicly detailed which specific practices it considers potentially unlawful, and neither company has issued a substantive response to the demands. That silence is typical early in an enforcement investigation, when companies are still negotiating the terms of production.

The timing matters because the Hugging Face hack has already sharpened scrutiny of security practices across the AI supply chain. But the commission's interest here appears to have been building independently of that incident, which suggests the investigation will not be resolved or shelved once the breach fades from headlines. Federal trade law gives the FTC broad authority over deceptive and unfair commercial practices, and AI vendors make a lot of public claims — about model capabilities, about safety testing, about how customer data is used — that fall squarely inside that authority.

For the wider field, the practical effect is that AI labs now face the same kind of document-intensive scrutiny that has reshaped other tech sectors. Startups and mid-sized model developers that were not named in the initial round could still be drawn in as the investigation widens, and outside counsel across the industry are already advising clients to preserve internal communications. The cost is not just legal fees: compelled testimony from executives can expose product decisions and internal debates that companies would rather keep private, and any resulting consent order could dictate how AI systems are described and sold.

What to watch next is whether the FTC converts these demands into a formal enforcement action or settles for commitments from the companies. The commission's recent history suggests it is willing to litigate, but AI cases raise novel questions about how consumer protection doctrine applies to probabilistic systems that can produce different answers to the same question. Another open question is whether other regulators, including state attorneys general and the European Union, follow with parallel inquiries. For now, the labs are gathering documents and preparing for depositions, and the answers they give under oath will shape the rules the rest of the industry eventually has to follow.