SGLang RCE Flaw Breached NEAR AI, Exposing Data From 18 Agents and ~100 Users
RedSide Security November 10, 2026 AI Security 4 views
An attacker exploited a remote code execution flaw in the open-source SGLang inference engine to compromise a NEAR AI server, potentially exposing requests — including those sent in confidential mode — and stealing data tied to 18 agents and about 100 users.
Continue reading: SGLang RCE Flaw Breached NEAR AI, Exposing Data Fr…
