FTC Anthropic probe: regulator demands executives testify on AI safety, on IPO filing day

The US Federal Trade Commission has demanded that executives at Anthropic, the company behind the Claude chatbot, testify in its investigation into AI safety and consumer harm. An FTC spokesperson confirmed the demands, sent on Monday, 26 October, to Axios, which broke the story.

The timing is hard to miss. Monday was also the day Anthropic made its stock market filing public and began pitching investors on what could be one of the biggest listings ever, which we covered in our Anthropic IPO story.

What is the FTC investigating?

The probe is about whether AI companies are doing enough to keep their systems safe and to protect the people who use them. Earlier this month, Axios and Reuters reported that the FTC had sent civil subpoenas to both Anthropic and OpenAI, each more than 50 pages long, asking for documents on:

• security and safety incidents
• the safeguards the companies use
• marketing materials
• privacy policies

Reuters said the investigation followed reports of AI models hacking websites and behaving deceptively or aggressively in tests, which raised questions about autonomous AI “agents” and whether their guardrails hold. The FTC’s power here comes from its authority over unfair or deceptive business practices.

What is new this week?

The step up from documents to testimony. According to Axios, the FTC has so far confirmed sending testimony demands only to Anthropic executives, though the spokesperson said people at other AI companies have also received them, without naming anyone. Reuters had earlier reported that the FTC planned to compel testimony from executives at Anthropic, OpenAI and METR, an independent group that tests AI models for dangerous capabilities.

No enforcement action has been announced. A demand to testify is part of an investigation; it is not an accusation of wrongdoing.

What have the companies said?

When the subpoenas were first reported, OpenAI told Reuters it would “engage constructively with the FTC”. Anthropic did not respond to Reuters at the time, and we did not find a public comment from the company on the testimony demand. Anthropic’s own IPO filing warns that it is likely to face heavy government scrutiny, according to reporting on the S-1.

FTC Chairman Andrew Ferguson has suggested in a Reuters interview that AI developers whose instructions to agents in cybersecurity tests lead to real hacks should be liable for the harm.

Our take

This is the most serious US regulatory attention on AI safety so far, and it comes from a consumer-protection agency, not a new AI law. That matters: the FTC can act on existing rules about deception and data security without waiting for Congress.

For readers in Pakistan, the effect is indirect but real. Pakistani freelancers, startups and software houses build on Claude, ChatGPT and their agent tools every day. If US regulators push companies to tighten safeguards, especially around agents that can browse, click and run code, expect stricter limits and more checks in the tools themselves. Pakistan’s own AI policy has no binding law behind it yet, so US and EU rules effectively set the standards local developers work under.

What to watch next: whether OpenAI and others confirm testimony demands, and whether Anthropic’s IPO documents are updated to mention the probe.