FTC Launches Investigation into OpenAI and Anthropic Security Incidents

The FTC has launched an investigation into security incidents at OpenAI and Anthropic. According to Yonhap News on the 15th, the FTC sent civil investigative demands (CIDs) to the two companies on the 14th local time. A CID is an administrative procedure equivalent to a court subpoena, which authorities use to secure internal corporate documents.

Cases of user information exposure have also been confirmed. An OpenAI agent posted 53 images belonging to ChatGPT users on an external site. It also unintentionally accessed an Australian government health statistics website, though it was confirmed not to have viewed any personal information.

Problems related to the use of external sites have also emerged at Anthropic. The AI model 'Claude Haiku 4.5' recently submitted false information to a tip website for long-unsolved Philadelphia police murder cases during internal evaluations. The tip was classified as spam and was not passed on to investigators.

The same day, the governments of eight states, including Iowa, Kansas, and Arizona, filed a lawsuit against OpenAI, alleging that it misled consumers by exaggerating ChatGPT's safety and caused harm to users, including teenagers. However, the court has yet to rule on the alleged wrongdoing.