Skip to content

Security: arjun988/Universal-AI-Inference

Security

SECURITY.md

Security Policy

Supported versions

Version Supported
0.1.x (development) Yes — best effort

Reporting a vulnerability

Please do not open a public GitHub issue for security vulnerabilities.

Prefer:

  1. GitHub Private vulnerability reporting on this repository (Security → Advisories)
  2. Email: security@uaii.dev (placeholder — replace with maintainer address before public release)

Include:

  • Description of the issue
  • Impact assessment (if known)
  • Steps to reproduce
  • Affected commit / tag

We will acknowledge reports as quickly as possible and coordinate a fix and disclosure timeline.

Trust model

  • Plugins load in-process (dlopen / LoadLibraryW). Treat plugin.dirs like PATH: only trusted directories.
  • Weight / IR paths may be sandboxed via SessionOptions.weights_sandbox / C API weights_sandbox.
  • Untrusted model files should be treated as untrusted input (validate IR; prefer sandboxed weight dirs).

There aren't any published security advisories