Threat watch / Breach
NEAR AI: an SGLang flaw on an inference server exposes about 100 users and 18 agents
AI platform NEAR AI says an attacker exploited a remote code execution flaw in SGLang, the open-source inference engine on its servers, on November 8, 2026. The attacker reached an internal system behind its deprecated Private Chat and took account details of about 100 users, plus chat history, settings and logs of 18 agents on a legacy deployment. NEAR AI says it fixed the flaw on all its inference servers and reported it to the SGLang project.
- Sector
- Tech & SaaS
- Way in
- Software flaw
- Product or vendor
- SGLang, NEAR AI
- How it happened
- NEAR AI says the attacker exploited a remote code execution vulnerability in SGLang, the open-source inference engine running on its servers, and from an inference server reached an internal system behind its deprecated Private Chat. No CVE or affected SGLang version has been published.[1][2] Checked 11 Nov 2026
- What was stolen
- Account details of about 100 users (email, name, and for some accounts plan and balance); chat history, settings and logs of 18 agents on a legacy NEAR AI Agents deployment that did not run in its trusted execution environment; and two NEAR AI Cloud customers' API keys found in logs, since revoked. NEAR AI says it saw no sign that stored Private Chat conversations or uploaded files were accessed, but the attacker could have observed requests processed on that server on November 8, including confidential-mode ones.[1][2] Checked 11 Nov 2026
- Who was impacted
- About 100 NEAR AI users, the owners of 18 agents on the legacy deployment, and two NEAR AI Cloud customers whose API keys were exposed.[1][2] Checked 11 Nov 2026
- Threat actor
- Not attributed
Timeline
- An attacker exploits a remote code execution flaw in SGLang on a NEAR AI inference server and reaches an internal system behind Private Chat. [1]
- NEAR AI publishes its incident report. It says it revoked sessions and admin credentials, replaced agent tokens, took Private Chat offline, deployed a fix for the SGLang flaw across all inference servers, rotated credentials, completed a forensic review and reported the flaw to the SGLang project. Access to NEAR AI Cloud is restored. [1]