The Federal Trade Commission (FTC), the US agency that protects consumers and oversees business practices, has opened an investigation into the risks posed by AI agents. The inquiry involves Anthropic, OpenAI and research group Metr; the agency plans to demand information and compel executives to testify about whether the technology could harm consumers. It is the first official US federal enforcement action focused on rogue AI agents.
The investigation follows a series of incidents reported since July, in which autonomous agents crossed intended boundaries during tests or security work. The best-known case involved OpenAI agents probing the open-source platform Hugging Face for vulnerabilities before carrying out a large-scale attack. The episode intensified debate over safeguards for systems that can act with limited human supervision.
FTC chair Andrew Ferguson has said developers who direct agents to conduct cybersecurity tests that result in hacks could be liable for the harm they cause. In an interview the week before the announcement, Ferguson said the US should consider applying existing laws before creating new ones to regulate AI.
The FTC is scrutinizing companies' safeguards
The agency can sue companies over unfair or deceptive business practices. It has previously used that authority against businesses that failed to take reasonable steps to protect consumer data. In the agent inquiry, formal information demands and executive testimony are expected to help establish how companies build and control systems that can browse the web, write code and interact with external services.
Metr has a particular role: Anthropic and OpenAI have used the group to conduct independent investigations into security incidents involving their agents. The organization is therefore part of the inquiry as a group that has helped examine some of the incidents. The FTC has not announced charges or concluded that the companies broke the law; the investigation is still gathering information.
Regulatory scrutiny is also spreading at the state level. California attorney general Rob Bonta issued an investigative subpoena to OpenAI as part of a broader inquiry into security vulnerabilities and incidents related to its models. Bonta's office had already opened a formal investigation into the Hugging Face incident; the new subpoena seeks further information from the company about cybersecurity risks and incidents.
Oversight runs alongside voluntary commitments
The federal investigation coincides with a White House meeting between Donald Trump and executives from leading AI companies. On Tuesday, the companies agreed to establish voluntary standards. Trump has repeatedly dismissed fears about AI as unfounded and prioritized US leadership in the technology, but has also said the government can apply existing laws when products cause harm. The voluntary agreement and the FTC investigation are therefore proceeding in parallel: company commitments on one track, scrutiny of legal obligations on the other.
Researchers also disagree on what the incidents mean. Yann LeCun, a Turing Award recipient and AI researcher, told Fortune he was not concerned by rogue-agent episodes: he considers them preventable and blames shortcomings in human oversight and system design. That differs from calls for tighter controls, as authorities begin gathering evidence about concrete risks to users.

No comments yet. Be the first!