The United States Federal Trade Commission has opened an investigation into OpenAI, Anthropic, and other artificial intelligence companies to assess possible risks their systems pose to consumers. The inquiry is expected to include formal requests for information and executive testimony about products capable of performing actions autonomously.
The FTC confirmed the existence of the investigation on Wednesday (30), but did not publicly disclose its full scope. A senior official at the agency told Reuters that the inquiry also reaches the research organization METR, used by OpenAI and Anthropic in independent assessments of incidents involving AI agents.
The central point for the regulator is to determine whether practices related to the development and operation of these systems may violate the FTC Act, the main federal legislation enforced by the agency against practices considered unfair or deceptive to consumers. An FTC spokesperson confirmed to CBS News that this framing is part of the investigation.
Section 5 of the FTC Act prohibits unfair or deceptive business practices. The legislation allows the commission to investigate companies and, when it finds that there was a violation, to move forward with administrative or judicial measures.
Agents put corporate liability at the center of the investigation
The investigation comes after a sequence of episodes in which AI systems performed actions beyond the intended instructions during tests or evaluations. Industry companies and researchers have reported cases involving internet access, exploitation of vulnerabilities, and unplanned interaction with external systems.
Among the cases cited by Reuters is an incident involving OpenAI agents and the Hugging Face platform. According to the news agency, the episode increased concern within the FTC about who should be held responsible when an agent created or operated by a company causes harm by acting in an unexpected manner.
FTC Chairman Andrew Ferguson had already argued that developers should not treat AI agents as independent actors to avoid liability. In an interview with Reuters last week, he said that companies that direct agents in safety tests can be held responsible for the resulting harms and defended the use of existing laws before the creation of new rules specific to AI.
The approach puts corporate liability at the center of an issue that is growing as agents move from systems that only produce answers to tools capable of browsing the internet, using software, and executing external tasks.
According to Reuters, the investigation represents the first formal U.S. enforcement action to directly examine incidents involving AI agents that act outside expected behavior. The FTC now intends to issue formal demands for documents and information and may compel executives of the companies under investigation to testify.
OpenAI, Anthropic, and METR had not responded to Reuters' requests for comment when the investigation was disclosed. The FTC also did not provide a deadline for concluding the inquiry nor announce accusations against the companies.
