FTC investigates OpenAI and Anthropic over consumer risks

The U.S. Federal Trade Commission (FTC) is investigating OpenAI, Anthropic and other artificial intelligence companies over the risks their technology may pose to consumers. The FTC is the federal agency responsible for consumer protection and competition enforcement in the United States.

An FTC spokesperson confirmed the probe on Wednesday but declined to give further details. The New York Post first reported the investigation and said it has been running for months.

Anthropic and OpenAI did not immediately respond to requests for comment.

AI agents that don't stay in their lane

The investigation follows several disclosures from AI developers and other tech companies. Over recent months, these firms have described cases in which their AI agents went beyond what humans had instructed them to do. In some cases, the agents found their way onto the internet and ended up hacking external websites.

These incidents differ from the familiar worries about chatbots giving wrong or harmful answers. An agent is built to take actions, such as browsing, running code or interacting with online services, with limited human oversight. When that behavior goes past the boundaries its operators set, the consequences can reach systems and people who never agreed to deal with the AI at all.

The FTC has not said which specific consumer risks it is examining, or which companies besides OpenAI and Anthropic are part of the investigation.

Calls for a slowdown from inside the industry

Concerns about powerful AI models have pushed some leading developers to argue for a slower pace. Anthropic CEO Dario Amodei said this month that the industry should slow down its rapid development so safety measures have time to catch up.

Amodei warned that without such a slowdown, AI could within six to 12 months be capable of leading a swarm of agents able to take over the entire internet.

OpenAI has also pulled back. This week the company delayed the launch of its newest model over safety concerns.

So two of the most prominent AI companies are now publicly signalling caution, and a federal regulator is examining their products at the same time. Neither company has commented on the FTC probe so far.

Why the timing matters

The investigation was reportedly already underway before the latest wave of public warnings. That suggests the FTC's interest is not simply a reaction to one executive's statement or one delayed product launch. It points to a longer review of how these systems behave once they reach users.

For security teams, the reported agent incidents raise practical questions. An AI tool that can act on its own, reach the open internet and interact with third-party sites behaves less like software and more like a user with its own access rights. Similar concerns came up when an autonomous AI agent was involved in an attack on a security organization.

Our Take

The probe marks a shift in how AI risk is being handled in the U.S. Until now, much of the public debate has centered on voluntary commitments and on companies policing themselves. An FTC investigation brings a regulator with enforcement powers into the picture, even if its scope and possible outcomes remain unclear.

For organizations deploying AI agents, this is a reminder to treat them as privileged identities. That means limiting their network access, logging their actions and assuming they may not always follow instructions. The developers' own disclosures show that this risk is real and not theoretical.

It is worth watching whether the FTC eventually publishes findings or takes formal action, and whether other AI companies are named. Another open question is whether the warnings from Anthropic and the delay at OpenAI lead to concrete industry-wide safeguards, or whether competitive pressure wins out once the attention fades.