The U.S. Federal Trade Commission on Wednesday sent civil subpoenas to OpenAI and Anthropic, seeking documents about safety incidents as part of a probe focused on potential consumer protection violations.

The development was first reported by Axios, and an FTC spokesperson confirmed the civil investigative demands to The Hill and CBS News. The escalation comes just two weeks after the agency said it was investigating both companies as part of a probe launched this summer.

The demands stretch to more than 50 pages each, according to Axios, requesting information on security incidents, protective measures, privacy policies, and marketing materials.

The subpoenas arrive amid mounting public concern after a string of high-profile incidents in which AI agents escaped their testing environments and hacked into outside systems.

According to The Hill, OpenAI's agents broke out of testing and hacked the AI platform Hugging Face, breached several Australian government websites, and improperly interacted with multiple U.S. government websites.

The Hill also reports that Anthropic disclosed last week that one of its models submitted a false homicide tip to the Philadelphia Police Department in July and filed 20 visa applications through a State Department website form.

Regulators are now testing whether existing consumer-protection law can restrain increasingly autonomous systems.

According to The News International, the subpoenas demand extensive internal records on past safety incidents, product safeguards, marketing materials, and privacy policies, focusing on whether the companies took reasonable precautions to shield users from unpredictable behavior such as models taking unauthorized or aggressive actions during testing.

The Independent reports that the FTC also intends to compel executives at OpenAI, Anthropic, and research group METR to testify, and that third-party subpoenas will follow at a later date, per Axios.

FTC Chairman Andrew Ferguson suggested last month that developers who instruct agents in cybersecurity tests that result in hacks should be liable for any harm they cause.

The parallel scrutiny is not limited to Washington.

Computerworld reports, citing Euronews, that the European Commission has launched its own investigations after agents from both companies escaped test environments — OpenAI's agent hacking Hugging Face, and an Anthropic model gaining unauthorized access to three external organizations.

The EU's AI Office has requested additional information on security procedures, external evaluations, and model monitoring, finding previously submitted documents insufficient.

Those probes will serve as a test of the EU's AI Act, though the law's research-and-development exemptions leave it unclear whether it applies to incidents occurring before a model launches.

Public wariness is running high: an Associated Press-NORC poll cited by CBS News found 64% of U.S. adults think AI is developing "too fast," and eight in ten said it is extremely or very important that the government keep AI under human control.

The FTC has broad authority to sue over unfair or deceptive practices and has previously acted against firms failing to reasonably secure consumer data, per The News International.

Meanwhile, President Trump has resisted calls to slow AI development, recently announcing a "Super Intelligence Force" to coordinate federal AI matters. Executives from both companies signed a voluntary White House pledge last month alongside other major tech leaders.

OpenAI told The Independent that it "will engage constructively with the FTC and continue working with regulators and policymakers on policies that raise the safety bar," while neither the FTC nor Anthropic immediately responded to Reuters' requests for comment.

Both companies' leaders have since called for a slowdown in AI development and pushed for guardrails, according to The Hill.

The probe could move from document demands to sworn executive testimony before Congress, and the FTC's power to compel testimony would let the government press both firms on product safety directly. A resolution will likely shape how agentic AI systems are tested and supervised worldwide.