NEAR AI says an attacker used an SGLang flaw to take user and AI agent data and possibly see confidential requests
NEAR AI · Incident report
Read the source (opens in a new tab)- What happened
- NEAR AI said on November 9, 2026 that an unauthorized party exploited a remote code execution flaw in SGLang, the open-source inference engine on its servers, and reached its infrastructure.
- Key points
- The attacker downloaded account details, including email addresses and names, for approximately 100 users and accessed 18 agents on a legacy deployment outside NEAR AI's confidential computing environment.
- On November 8, the attacker could also have observed user requests processed on the breached server, including requests running in confidential mode.
- NEAR AI says confidential computing "does not protect against a vulnerability in the software running inside the environment," and it has reported the flaw to the SGLang project.
- Two Cloud customers had API keys exposed in logs, since revoked; NEAR AI found no sign that stored Private Chat conversations or uploaded files were accessed.
- Evidence
- Company incident report on its own systems, written after a forensic review. For the 18 agents, NEAR AI says the data taken included chat history, settings and logs.
- Uncertain
- NEAR AI says the flaw let the attacker "access requests in processing on that server, including confidential ones," but does not say whether any specific requests were seen. It does not name the attacker.