SGLang RCE breach at NEAR AI reaches 18 agents and about 100 users
NEAR AI says an attacker exploited a remote code execution flaw in SGLang, reached 18 AI agents, and downloaded account details for about 100 users.
NEAR AI says an attacker exploited a remote code execution (RCE) flaw in SGLang, an open-source inference engine, and ran code on one of its inference servers. On November 8, the attacker could have observed requests being processed on that server, including requests sent in confidential mode.
What is affected
The attacker reached 18 AI agents on an older deployment of NEAR AI Agents that did not run inside the company’s trusted execution environment (TEE), taking chat history, settings and logs. The attacker also downloaded account details for approximately 100 users from an internal system behind the deprecated Private Chat product; the data included email addresses, names, and for some accounts plan and balance information. Two customers’ NEAR AI Cloud application programming interface (API) keys were exposed in logs, and NEAR AI revoked both.
Why it matters
NEAR AI says a TEE does not protect against a vulnerability in the software running inside the environment. The accessed server could have exposed confidential-mode requests, and the 18 agents ran on a deployment outside the TEE.
Response and action
NEAR AI deployed a fix for the SGLang flaw on all inference servers and restored access to NEAR AI Cloud. It revoked all active sessions and administrator credentials, replaced all agent tokens, and took Private Chat offline. It also reported the SGLang flaw to the SGLang project. According to The Hacker News, NEAR AI has not identified the flaw by CVE (Common Vulnerabilities and Exposures) number or named the affected SGLang versions.
No fixed SGLang release is named in the cited sources. NEAR AI’s response included a fix across its inference servers and a report to the SGLang project.
Sources
Related reading
3 postsDiscussion
Spotted an error? Send corrections to [email protected] .