NEAR AI says SGLang vulnerability exposed user account details in breach

The significance here lies less in the breach itself than in where it originated: a flaw in the SGLang serving framework, an open-source inference component, rather than in NEAR AI's own application layer. That shifts the exposure question toward every other project running the same serving stack, since the disclosure says nothing about whether the flaw was patched upstream or how widely it is deployed. NEAR AI's assurance that Private Chat content and uploads were untouched narrows the damage to account metadata, but the absence of affected-account counts and timing leaves severity unresolved. Whether other SGLang-dependent deployments disclose similar exposure is the open question.