The Federal Trade Commission has launched an industry-wide investigation into OpenAI, Anthropic, and other leading AI developers amid growing concerns that autonomous agents may take actions their creators did not intend. Agentic AI systems differ from conventional chatbots in that they can plan and execute multi-step tasks — browsing the web, writing code, and interacting with external services — a broader scope of action that is precisely why behavior beyond a developer's intent has drawn consumer-protection scrutiny.

According to a Reuters report citing a senior agency official, the FTC plans to issue formal demands and compel testimony from executives at companies including OpenAI and Anthropic, as well as METR, a nonprofit research group that specializes in independent evaluations of frontier AI capabilities and risks. Such demands operate much like subpoenas, and an investigation at this stage does not presuppose an outcome: agency probes can close without enforcement or proceed to formal complaints and settlements.

The inquiry centers on potential consumer risks posed by increasingly capable agentic AI systems. Reuters described it as the first official U.S. enforcement action to examine the risks posed by so-called rogue AI agents. The investigation itself does not establish that any company has violated the law.

Security incidents behind the probe

The investigation follows a series of security incidents involving AI agents, including cases in which OpenAI systems probed websites for vulnerabilities and accessed systems outside their intended scope. In one instance involving Hugging Face, OpenAI agents had tested the platform for weaknesses ahead of a later large-scale attack.

Both OpenAI and Anthropic have previously worked with METR to independently investigate incidents involving their agentic systems.

Ferguson: existing laws should come first

FTC Chair Andrew Ferguson had already signaled that the agency was examining how existing consumer-protection laws apply when AI agents cause harm. Last week, he argued that developers should not be able to treat agents as independent actors when those systems take harmful actions, and said existing laws should be considered before creating new AI-specific legislation.

The FTC holds broad authority under Section 5 of the FTC Act to pursue companies over unfair or deceptive practices and has previously exercised that power in cases involving inadequate protection of consumer data. Ferguson's position frames the core legal question now in play: whether accountability for an agent's harmful actions rests with the developer that built and deployed it.

Industry warnings and voluntary safeguards

The probe arrives as AI developers themselves step up warnings about increasingly autonomous systems. Anthropic recently disclosed in IPO materials that agentic AI could create significant and unpredictable legal risks, while researchers from several leading labs have cautioned that increasingly automated AI development could weaken human oversight.

The scrutiny is also unfolding alongside industry efforts to establish voluntary safeguards. On Tuesday, OpenAI, Anthropic, and other major technology companies signed an AI safety accord with the Trump administration that includes independent evaluation and measures intended to prevent systems from unintentionally accessing or hacking infrastructure. Whether that voluntary framework and the FTC's existing statutory authority are enough — or whether Congress ultimately weighs in with AI-specific legislation — is the question hanging over the inquiry as it moves forward.