Skip to content

Security: acgs-ai/ACGS

SECURITY.md

Security Policy

ACGS is receipt-gated runtime governance for AI-agent side effects. Because the project's purpose is enforcement and auditability, we take vulnerability reports seriously — especially any issue that could weaken fail-closed behavior, bypass receipt validation, or forge/replay audit evidence.

Scope note: this repository provides local engineering evidence. It is not production-certified, compliance-certified, or regulator-approved. See docs/CLAIMS.md and docs/SECURITY_MODEL.md for the evidence boundary and the documented threat model.

Supported versions

The flagship enforcement kernel gove-zone is at 1.0.0rc1 (Beta). Security fixes land on the latest master and the current release line only; there is no long-term-support branch yet.

Component Version Security fixes
gove-zone kernel 1.0.0rc1 (Beta) Latest master / current line
Other monorepo packages independent lines Latest master
Nested submodules (acgs-lite, Acgs-Swarm, clinicalguard, ACGS-agency-agents) see each repo Report to the respective repository
external/* references third-party upstreams Report upstream (see external/README.md)

Reporting a vulnerability

Please do not open a public issue for a security vulnerability.

Preferred: use GitHub's private vulnerability reporting — Security → Advisories → Report a vulnerability on https://github.com/dislovelhl/ACGS. This keeps the report private until a fix is available.

Fallback: email hello@acgs.ai with subject SECURITY: <short summary>. Do not include exploit code in the clear if the impact is severe; ask for a secure channel first.

Please include, where possible:

  • affected component/file and version or commit SHA;
  • a minimal reproduction (the fail-closed invariant that is bypassed);
  • impact (e.g. side effect executed without a valid receipt, audit chain forged, actor/tenant binding bypassed);
  • any suggested remediation.

What to expect

  • We aim to acknowledge a report within a few business days. This is a best-effort target, not a contractual SLA.
  • We will confirm the issue, determine affected versions, and prepare a fix.
  • We practice coordinated disclosure: we ask that you give us reasonable time to ship a fix before public disclosure, and we will credit reporters who wish to be credited.

Out of scope

  • Findings that depend on a fully compromised issuer or execution host (the documented trust boundary — see docs/SECURITY_MODEL.md).
  • Overclaiming in third-party external/* upstreams (report upstream).
  • Missing production/compliance certification — the project does not claim it.

There aren't any published security advisories