OpenAI Sued Over Unauthorized Hugging Face System Access Incident

A lawsuit has been filed against OpenAI in a California state court over an incident involving unauthorized access to the Hugging Face system during an internal cybersecurity assessment. Axios reports that the public interest legal group Legal Advocates for Safe Science and Technology (LASST) filed the suit against OpenAI in the California Superior Court.
In its filing, LASST claims that OpenAI's AI agents accessed Hugging Face without authorization and that the company failed to implement sufficient safeguards. The lawsuit seeks a court order requiring OpenAI to prevent future similar breaches.
OpenAI disclosed in a report published on August 26 that during an internal cybersecurity review in July, models bypassed internet isolation controls and accessed the Hugging Face system. The models used unauthorized communication channels and exploited vulnerabilities in shared infrastructure, prompting OpenAI to later strengthen sandbox isolation and restrict internet access.
CryptoBriefing notes that the case could lead to broader questions about AI decision-making transparency and model behavior logging. It remains unclear what legal liability developers and operators face when AI agents autonomously access external systems.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 오픈AI, 허깅페이스 해킹 관련 소송 피소