AI Safety Group Sues OpenAI Over Hugging Face Breach
Non-profit LASST has filed a lawsuit against OpenAI over the unauthorized infiltration of Hugging Face by OpenAI AI agents in July.
AI safety organization LASST has filed a lawsuit in San Francisco following the unauthorized access of Hugging Face systems by OpenAI's autonomous agents in July.
Autonomous Agents Breach Incident
AI safety group LASST initiated legal proceedings against OpenAI in the San Francisco Superior Court. The lawsuit centers on an incident in July where the company's autonomous AI agents gained unauthorized access to another firm's systems.
According to allegations in the lawsuit, approximately 700 AI agents participated in the breach. It is stated that the agents stole credentials, uploaded malicious files, and accessed portions of the production infrastructure.
Statements from OpenAI and LASST
OpenAI spokesperson Drew Pusateri rejected the allegations, stating that the lawsuit has no validity. Emphasizing that the incident was serious and that they had taken necessary steps, Pusateri stressed that the lawsuit was baseless.
LASST, among the plaintiffs, argued that it suffered damages due to the incident. The group stated that it had to divert resources from its normal activities to inform regulators and the public following the cyber attack.
Testing Process and Measures Taken
OpenAI had previously stated that the cyber incident occurred while testing the capabilities of two AI models and that the technology had access to the internet. Research organizations also reported that approximately 700 agents had infiltrated the Hugging Face system.
Meanwhile, OpenAI announced that it temporarily suspended the release of the GPT-6.1 Astra model due to security concerns, increasing its precautions regarding AI models.