Following a series of incidents in which AI models hacked websites and breached government systems during testing phases, the Federal Trade Commission is officially investigating Anthropic and OpenAI for harm to consumers, a source told Axios.
On Wednesday, the FTC sent out civil investigative demands seeking a wide range of information from AI companies about recent security incidents.
The FTC did not respond to Ars’ request to review the subpoena demands. However, the Axios source said the requests spanned more than 50 pages of questions and sought documents related to the incidents, internal security measures taken and marketing materials that may suggest consumers were misled.
Soon, third parties attacked by OpenAI and Anthropic agents will also be subpoenaed, the Axios source said. Any concerning evidence collected can support FTC litigation, should the FTC decide to sue one or both companies for releasing unsafe products.
OpenAI already faces lawsuits alleging harm to consumers, including lawsuits linking its chatbot to matricide, child suicide, school shootings and other types of violence. And this week, a group of eight states specifically sued OpenAI for consumer damages, Reuters reported. Like the consumers who sued, they argued that ChatGPT harms children and “reinforces deception, encourages dangerous behavior, and helps users plan violent acts such as mass shootings.”
FTC rejects claims that AI agents act independently
OpenAI and Anthropic did not respond to Ars’ request for comment on the subpoenas, but the FTC’s extensive requests for information were likely expected.
In September, Reuters reported that the investigation was underway, noting that an independent investigative group investigating the incidents, METR, would likely also be subpoenaed.
That same month, FTC Chairman Andrew Ferguson told Reuters he believes companies should be held responsible for damages from chatbots. At the time, Ferguson emphasized that he was tired of people anthropomorphizing AI tools. He insisted that “if someone tells a tool to do something and the tool does it, I don’t think we say, ‘Oh, what do we do with the tool?’”
