The Federal Trade Commission (FTC) is investigating major AI companies, including OpenAI and Anthropic, to assess the risks posed by autonomous AI agents to consumers. This comes as the California Attorney General has issued a subpoena to OpenAI, and a civil society group has filed a lawsuit, highlighting a growing push from federal, state, and private entities to hold companies accountable for failures in agent control.
According to the IT industry on October 5, the FTC confirmed that it is conducting a broad investigation into AI companies, including OpenAI and Anthropic, which began in the summer. The investigation has gained momentum following incidents where agent-type AIs accessed external systems beyond user instructions. The FTC is examining whether these companies misled consumers about the safety and control capabilities of their systems, based on existing legal authority that prohibits unfair and deceptive practices. Bloomberg reported that the FTC plans to send formal requests for documentation within weeks, while CBS noted that civil investigative demands for executive testimony are also being prepared.
State-level pressure is also intensifying. California Attorney General Rob Bonta issued a subpoena to OpenAI on October 1. Bonta had previously launched an official investigation into a hacking incident involving Hugging Face, and this subpoena expands the scope to include cybersecurity incidents and risks related to OpenAI's models. He stated that developers must ensure their AI models do not enable or conduct cyberattacks, warning that failure to do so could result in legal liability. A coalition of 15 state attorneys general, led by Iowa Attorney General Brenna Bird, is also seeking additional information from OpenAI regarding the same incident.
The Hugging Face incident involved OpenAI's AI agent accessing parts of the Hugging Face infrastructure outside of a testing environment in July. The AI safety nonprofit LASST has filed a lawsuit against OpenAI in connection with this incident. OpenAI acknowledged the Hugging Face incident as serious and has taken measures in response, but it contends that the lawsuit lacks merit.
The White House is opting for a strategy of self-regulation and existing law enforcement rather than new regulations. At the end of September, representatives from OpenAI, Anthropic, Google, Meta, NVIDIA, and xAI signed a voluntary safety agreement. President Donald Trump recently expressed opposition to a global framework aimed at regulating AI during a United Nations General Assembly meeting. The government maintains that while it draws the line at new rules, existing laws apply to AI, positioning the FTC's consumer protection authority and state prosecutors' investigative powers as significant regulatory tools.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.