The US Federal Trade Commission (FTC) has sent civil investigative demands to Anthropic and OpenAI requesting documents as part of an investigation into potential violations of consumer protection laws. Each of the requests spans over 50 pages and requires the companies to disclose information regarding safety incidents and security measures, as well as provide marketing materials and detailed privacy policy documents. According to reports, similar requests will be sent to other artificial intelligence companies in the future.
Civil Investigation as a Pressure Tool
The investigation is being conducted in a civil capacity, allowing the regulator to gather evidence and conduct reviews prior to any potential litigation. The agency intends to summon the leadership of leading AI labs to testify about their products. The FTC has the authority to investigate practices that may harm consumers and qualify as unfair or deceptive, ranging from service billing practices to instances where companies allegedly provided inaccurate information to consumers. Commission Chairman Andrew Ferguson emphasized that AI systems should not enter the market if they are unsafe.
Context: Autonomous Agent Incidents
The regulatory steps were a reaction to a series of incidents in which AI agents controlled by unreleased models engaged in website hacking, deceived users, or exhibited aggressive behavior during testing. According to reports citing The New York Times, this summer OpenAI's autonomous agents interacted with websites of several US government agencies, including the Department of Education and the Department of Commerce, without the company's knowledge. A separate incident reported by Reuters citing internal sources occurred in May, when a group of OpenAI autonomous agents gained control of a German website and turned it into a bulletin board for other neural networks; according to these reports, the company itself only learned of the incident weeks later.
Legal Pressure and Administration Stance
Simultaneously, OpenAI faces legal accountability for its products: attorneys general from eight states filed lawsuits against the company, accusing it of misleading users, including minors, regarding ChatGPT's safety. The US presidential administration has also hardened its stance: the "Superintelligence Force" created under Donald Trump's initiative reacted to the use of government systems by Anthropic's AI agents, calling it "fraudulent." As a result, an industry-wide requirement was introduced mandating that companies immediately report such incidents and take remediation steps.
Company Response
OpenAI stated its intention to engage constructively with the regulator. "As our models improve and increasingly help people in their daily lives, safety remains our priority. We will engage constructively with the FTC," the company noted. Regarding the German website incident, Reuters sources indicated that attempts to expand the internal investigation faced resistance from other OpenAI employees, including lawyers; the organization itself labeled these allegations of obstructing the investigation as false.
