Arbitrary Code Execution Vulnerability in FastGPT by Labring
CVE-2026-33075

9.4CRITICAL

Key Information:

Vendor

Labring

Status
Fastgpt
Vendor
CVE Published:
20 March 2026

What is CVE-2026-33075?

FastGPT, an AI Agent building platform developed by Labring, has a vulnerability in the fastgpt-preview-image.yml workflow found in versions 4.14.8.3 and earlier. The issue allows unauthorized execution of arbitrary code and exfiltration of secrets by external contributors. This occurs because the workflow uses pull_request_target, granting it access to repository secrets while incorporating code from the contributor's fork. Furthermore, the attacker can utilize their own Dockerfiles to build and push images, leading to potential supply chain attacks through the production container registry. A patch has not yet been released for this vulnerability.

Affected Version(s)

FastGPT <= 4.14.8.3

References

https://github.com/labring/FastGPT/security/advisories/GH...
x_refsource_CONFIRM

CVSS V4

Score:
9.4
Severity:
CRITICAL
Confidentiality:
High
Integrity:
High
Availability:
High
Attack Vector:
Network
Attack Complexity:
Low
Attack Required:
None
Privileges Required:
Undefined
User Interaction:
None

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.