Executive Summary
A remote code execution exploit in the SGLang inference engine compromised a NEAR AI inference server and may have exposed requests processed there. AI platform owners, agent operators, and teams handling confidential inference requests should assess similar dependency and isolation risks.
What Happened
An attacker exploited a remote code execution flaw in SGLang, an open-source inference engine, and ran code on one of NEAR AI's inference servers last Sunday. The attacker could also have seen requests being processed on that server that day, including requests sent in confidential mode, the company said earlier this week. That mode processes requests inside a trusted execution environment (TEE
Affected Products
- SGLang
- NEAR AI
Vulnerability / Attack Technique
Remote code execution
Technical Mechanism
The attacker exploited a remote code execution flaw in SGLang, an open-source inference engine, to run code on an inference server. The compromised server processed requests, including requests sent in confidential mode inside a trusted execution environment, and the attacker could also have seen those requests.
AI Security Category
LLM03:2025 Supply Chain
Part of the AI Supply Chain and AI SBOM topic.
Relevant Frameworks
RealGround Analysis
This incident shows how a vulnerability in an open-source inference component can affect applications and agents that depend on it. Teams running AI workloads should maintain an inventory of inference engines and agent dependencies, validate isolation boundaries, and verify that confidential processing limits exposure after host compromise.
Recommended Controls
- Maintain a software bill of materials for inference engines and AI agent dependencies, including SGLang versions.
- Apply security patches and monitor advisories for open-source inference components.
- Isolate inference servers from unrelated workloads and restrict their network access.
- Test trusted execution environment configurations against host-level compromise and request exposure.
- Log and review inference-server execution, access, and request-handling activity for signs of compromise.
Related RealGround Service
AI Agent SBOM and LLM Mapping · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/11/sglang-rce-exploit-breaches-near-ai.html
Summary and analysis written by RealGround from the source above; last updated 2026-11-11.
