Package vulnerability, severity high
SGLang RCE flaw exploited to breach NEAR AI inference server; data of 18 agents and ~100 users stolen
Published
An attacker exploited a remote code execution flaw in the open-source inference engine SGLang and ran code on one of NEAR AI's inference servers. The attacker could also have seen requests processed on that server that day, including confidential-mode requests. The article reports data stolen from 18 agents and about 100 users.
What's affected
- sglang (pypi)every version
What to do
- Check whether you run SGLang inference servers and apply the maintainers' fix for the RCE flaw
- If you use NEAR AI, review what requests you sent to the affected server around the incident
Check whether you were affected
Pharon tells teams automatically when this affects something they run. To check a machine, run this in Claude Code with the Pharon plugin, in the repository that uses it:
/pharon:check-incident sglang-rce-near-ai-breach-2026-11References
Notice sglang-rce-near-ai-breach-2026-11, signed with key feed-prod-2026-10 (Ed25519). The Pharon CLI verifies the signature before using a notice.