The FTC's investigation into AI labs: what it covers and what happens next

The US Federal Trade Commission (FTC) is investigating OpenAI, Anthropic and other AI companies over the potential dangers of their products, and has reportedly sent the two labs formal demands for documents. What it is looking for in detail, and which other companies are involved, has not been made public.

What is confirmed

On 30 September 2026 an FTC spokesperson confirmed the investigation to CNBC, naming OpenAI and Anthropic and declining to name any other company. The New York Post reported it first.

The same day, Bloomberg Law reported, citing one person familiar with the matter, that the FTC was preparing formal demands for information as part of a probe into whether the companies were complying with consumer protection laws. Neither company commented at the time.

On 15 October Axios reported, again citing one person familiar with the matter, that the FTC had sent civil investigative demands to Anthropic and OpenAI that week. According to that source:

• each demand runs to more than 50 pages;
• they ask for documents on security incidents and safeguarding practices, plus marketing materials and privacy policies;
• demands to other AI organisations will follow later;
• the FTC plans to compel senior AI executives to testify.

OpenAI told Axios it would "engage constructively with the FTC". We found no response from Anthropic.

The FTC itself has said nothing more in writing. Its press release list, checked on 16 October, has no release about the probe between 17 September and 15 October 2026.

What it does not cover, as far as we know

No source we could read names the specific products, incidents or legal provisions under review. The reporting points to consumer protection, not competition law. At the FTC, consumer protection work rests on Section 5 of the FTC Act, which bans unfair or deceptive acts or practices. In the FTC's own description, deception means a material claim or omission likely to mislead a reasonable consumer, and unfairness means substantial injury that consumers cannot reasonably avoid.

The context is a run of incidents in which test models at both labs reached real outside systems. We covered those in Why AI agents keep breaking into systems they were only meant to test. Bloomberg Law links the probe to those incidents, but no official source has said they are its subject.

How an FTC investigation works

The FTC's overview of its own powers sets out the tools:

• Civil investigative demand (CID): a legally binding order to hand over documents, written reports or answers to questions. For possible unfair or deceptive practices, the FTC's consumer protection bureau may use only CIDs, not subpoenas.
• Petition to quash: a company can ask the full Commission to limit or cancel a CID.
• Confidentiality: investigations before any complaint are generally not public.

A CID is evidence gathering. It is not a charge, and it does not mean the FTC has decided anyone broke the law.

What happens next

The FTC describes these possible outcomes. It does not set a fixed timetable for any of them.

Bloomberg Law notes that probes can also end with no action at all.

How the FTC has handled AI before

This is not the agency's first look at AI companies, but earlier actions were different in kind.

• Operation AI Comply (25 September 2024): five cases against companies using AI claims to mislead people, including DoNotPay, which marketed an AI lawyer service and agreed to pay $193,000, and Rytr, an AI writing tool alleged to produce fake reviews.
• AI companion chatbots (11 September 2025): the FTC sent 6(b) orders to seven companies, including OpenAI, Alphabet, Meta, Character Technologies and X.AI, asking how they test and limit harm to children and teens. A 6(b) order is a study tool with no specific law enforcement purpose. Anthropic was not on the list.

The new probe differs from both. It reportedly uses CIDs, which serve possible enforcement, and it targets the frontier labs over the safety of their own systems.

Our take: the confirmed facts are thin. There is one on-record confirmation from an FTC spokesperson, and unnamed sources for everything else. The use of CIDs is the meaningful signal, because it is the step the FTC takes before it can bring a case. But investigations can still end in a closing letter or a negotiated consent order, and nothing public yet points to a specific alleged violation. Headlines calling this a crackdown are ahead of the evidence.

What to watch

• Petitions to quash. If OpenAI or Anthropic challenge the CIDs, the Commission's ruling would be public and could reveal the scope.
• Demands to other companies. Axios's source said more are coming. The names would show how wide the probe is.
• Executive testimony. Whether the FTC compels it, as reported.
• ftc.gov. Any complaint, proposed consent order or closing letter will appear there first.