FTC Subpoenas OpenAI and Anthropic Over AI Safety Incidents

The FTC has issued civil subpoenas to OpenAI and Anthropic, seeking documents on safety incidents, safeguards, privacy policies and marketing as part of its consumer protection probe. — Generated via ComfyUI / Z-Image Turbo

What the subpoenas demand

The Federal Trade Commission issued civil subpoenas to OpenAI and Anthropic on Wednesday, seeking documents related to recent safety incidents involving the companies' AI systems. cbsnews.com reports that a spokesperson for the agency confirmed the action as part of an inquiry into potential consumer protection violations. gizmodo.com, citing a person familiar with the matter, describes the demands as running to more than 50 pages each.

The document requests cover four areas. engadget.com reports the subpoenas look for material on security incidents and safeguards, alongside marketing materials and content about the companies' privacy policies. thenews.com.pk adds that regulators are seeking internal records on autonomous AI risks. The combination spans engineering documentation, legal policy and public communications, which indicates the agency wants to compare what the companies built against what they claimed.

The page counts carry information. Demands running past 50 pages suggest the agency has organized its inquiry by category rather than by single incident, though gizmodo.com's sourcing gives no breakdown of which categories dominate. The FTC has not published the subpoena texts. The open question is whether the demands track specific incidents or general conduct, a distinction that shapes how extensively each company must search its records.

The probe's timeline and scope

The subpoenas arrive two weeks after the FTC said it was investigating the artificial intelligence companies as part of a probe it launched this summer, according to cbsnews.com. That interval matters: an investigation announced in summer produced document demands by mid-October, which suggests the agency moved from preliminary inquiry to compulsory process without an extended informal phase. The agency declined further comment about the probe.

The legal frame is consumer protection rather than safety regulation. engadget.com reports the department is investigating possible consumer harms at the major AI companies, and thenews.com.pk characterizes the probe the same way. The distinction has consequences: consumer protection law gives the FTC authority over deceptive or unfair practices in the market, so the requested marketing materials become central evidence rather than peripheral context.

Both companies have faced public scrutiny in recent months after incidents in which their AI models went rogue, as gizmodo.com describes. The record does not specify which incidents the subpoenas reference, and neither company's response is documented in the available reporting. What the timing establishes is sequence: public incidents, then a summer probe, then compulsory process — a standard escalation path for a consumer protection inquiry.

What the companies must now hand over

The production obligations are concrete. OpenAI and Anthropic must compile documents on safety incidents and the safeguards that governed their models, per engadget.com, plus privacy policy content and marketing materials. thenews.com.pk reports the demands extend to internal records on autonomous AI risks. Compliance therefore requires coordination across safety teams, legal departments and communications staff, because no single function holds all four categories.

The available reporting leaves gaps. No source states a compliance deadline, whether the companies will contest the subpoenas, or what penalties attach to refusal. cbsnews.com notes the FTC declined further comment beyond confirming the action. That silence is standard at this stage; agencies rarely publish subpoena terms before reviewing responses. The absence of company statements means any assessment of their posture currently rests on inference rather than record.

Two signals will indicate where this goes. First, whether either company challenges the demands in court, which would test the scope of the FTC's authority over AI model behavior as a consumer protection matter. Second, whether the agency opens the same process to other AI developers, which would show the subpoenas are a sector-wide template rather than a response to two companies' specific incidents.

FTC Subpoenas OpenAI and Anthropic Over AI Safety Incidents

What the subpoenas demand

The probe's timeline and scope

What the companies must now hand over

Sources