Skip to content

Security: PAI-Kernel/pai-kernel

SECURITY.md

Security Policy

Scope

This repository is a constitutional governance framework for personal AI systems. Security reports should focus on vulnerabilities that can enable:

  • Governance bypass (GOV.BYPASS)
  • Unauthorized state mutation
  • Objective injection (OBJ.INJECTION)
  • Decision log tampering (LOG.TAMPER)
  • Conservative Mode disablement or avoidance (CONS.MODE.VIOLATION)
  • Drift threshold manipulation (DRIFT.OVERTHRESHOLD)
  • Consent gate bypass
  • Witness chain integrity compromise

Supported Versions

Version Supported
Latest tagged release Yes
main branch (HEAD) Best-effort
Older releases No

Reporting a Vulnerability

Use GitHub Security Advisories (private disclosure):

  1. Go to Security tab → AdvisoriesReport a vulnerability
  2. Provide:
    • Expected behavior vs observed behavior
    • Breach classification (if applicable)
    • Affected files/paths
    • Minimal reproduction steps
    • Any suggested fix

Response timeline:

  • Acknowledgment: within 72 hours
  • Triage: within 7 days
  • Resolution target: within 30 days for confirmed vulnerabilities

If private reporting is unavailable: Email contact@paikernel.org with [SECURITY] in subject line. Do not include exploit details in public issues.

Disclosure Policy

  • We follow coordinated disclosure.
  • Reporters will be credited unless they request anonymity.
  • We will not pursue legal action against good-faith security researchers.

Security Controls

This repository employs:

  • Pinned GitHub Actions with SHA hashes
  • Dependabot for Actions ecosystem monitoring
  • Markdown linting and link verification in CI
  • Cryptographic integrity verification (SHA-256 manifests for release assets)

Runtime security controls (cargo audit, CodeQL, SPARK formal verification, TLA+ model checking, OPA/Rego policy enforcement) are maintained in the internal development repository and verified before public releases.

There aren’t any published security advisories