SGLang RCE Exploit Breaches NEAR AI, Steals Data From 18 Agents and About 100 Users
Read original article ↗AI Summary
An attacker exploited a remote code execution vulnerability in the open-source SGLang inference engine to breach NEAR AI's infrastructure, gaining access to an inference server running within a trusted execution environment (TEE). Despite the TEE's design to protect confidential data, the flaw in SGLang allowed the attacker to access processed requests, including those in confidential mode. The attacker exfiltrated account details of about 100 users from a deprecated internal system called Private Chat and accessed data from 18 AI agents on an older non-TEE deployment, including chat history, settings, logs, and two exposed NEAR AI Cloud API keys, which have since been revoked.
AI-extracted · verify before operational use
No entities or IoCs were extracted from this article.