Near AI Confirms Security Breach Affecting 100 Users and 18 Agents

Near AI disclosed a security breach on Nov. 8, 2026, that exposed account information for roughly 100 users and compromised data from 18 legacy AI agents after an intruder exploited a vulnerability in an open-source inference engine. The flaw, tied to SGLang, potentially exposed confidential AI requests routed through Near AI's infrastructure. The incident matters because it highlights the security risks facing decentralized AI platforms that rely on third-party open-source components to serve user inference requests. What to watch next is whether Near AI patches the inference engine dependency, discloses the full scope of affected accounts, and whether similar SGLang-based deployments across the ecosystem face the same exposure.