FastGPT, a knowledge-based AI application platform by labring, has a critical privilege assignment vulnerability (CVE-2026-50562) present in commit 22ebfacbb43311e9b73294040ae0eb87390c6bba and earlier. The vulnerability involves GitHub Actions workflows where privileged workflow_run jobs (.github/workflows/preview-docs-push.yml and preview-fastgpt-push.yml) can download artifacts created from untrusted pull request code in preview-docs-build.yml and preview-fastgpt-build.yml. This flaw enables an attacker to push malicious Docker images to GHCR and deploy documentation previews with access to secrets such as KUBE_CONFIG_CN. The vulnerability is associated with CWE-266 (Incorrect Privilege Assignment), CWE-494 (Download of Code Without Integrity Check), and CWE-829 (Inclusion of Functionality from Untrusted Control Sphere). There is no vendor advisory or patch available at this time.

An attacker can exploit this vulnerability to push attacker-controlled Docker images to the GitHub Container Registry and deploy them with access to sensitive secrets, potentially leading to unauthorized code execution and compromise of deployment environments. The vulnerability allows privilege escalation within GitHub Actions workflows by leveraging untrusted code artifacts, posing a high risk to the confidentiality and integrity of the affected systems.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is released, it is recommended to review and restrict the privileges of GitHub Actions workflows, especially those that handle untrusted pull request code. Avoid allowing privileged jobs to consume artifacts from untrusted sources and consider disabling or isolating workflows that build and deploy from untrusted code. Monitor for updates from labring regarding official patches or mitigations.