Near AI disclosed that an attacker exploited a remote code execution vulnerability in SGLang, the open-source engine used to run its AI models. The intrusion allowed the attacker to download the names and email addresses of about 100 users, along with subscription and account balance information. The attacker also accessed 18 AI agents in an older deployment, retrieving their chat histories, settings and logs. Two API keys belonging to Near AI Cloud customers were exposed and revoked, while confidential requests processed on November 8 may have been visible, with no evidence that stored Private Chat conversations were accessed.
Source: Read the original article


