CVE-2026-73678: MindsDB Minds Platform unauthenticated remote code execution in agent scratchpad

CVE-2026-73678 · Severity: critical · CVSS 10 · Published 2026-08-14

MindsDB Minds Platform is a desktop application that provides AI-powered data analysis and workflow automation. An unauthenticated remote code execution vulnerability allows attackers to execute arbitrary operating system commands by submitting crafted prompts to an unprotected API endpoint. This enables attackers to steal SSH keys, stored credentials, and environment secrets with full system privileges, even without knowing the victim's credentials.

Quick answers

Is CVE-2026-73678 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-73678 being exploited in the wild.
How severe is CVE-2026-73678?
Critical severity, with a CVSS score of 10, published 14 August 2026.
Which products and versions are affected?
MindsDB Minds Platform 26.1.0 and earlier.
Is there a fix for CVE-2026-73678?
No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.
How is CVE-2026-73678 attacked?
It is attacked over the network or with local access.
Am I affected by CVE-2026-73678?
Check whether you run MindsDB Minds Platform at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

Technical details

The vulnerability is a remote code execution flaw in the Anton agent's scratchpad tool within the cowork-server FastAPI backend. The root causes are: (1) no authentication middleware protecting the entire /api/v1/ API router, (2) CORS wildcard configuration allowing cross-origin browser requests, and (3) unrestricted exec() call on LLM-generated Python code without sandboxing. An attacker can inject their own LLM API key via the unauthenticated PUT /api/v1/settings/ endpoint, then POST a crafted prompt to /api/v1/responses/ (also unauthenticated) that directs the Anton agent to invoke the scratchpad tool with arbitrary Python code. The compiled code is executed with exec() inside the server process namespace, achieving full OS command execution as the user running the application. No patch is currently available for versions 26.1.0 and earlier.

Affected products

  • MindsDB Minds Platform 26.1.0 and earlier

Timeline

  • 2026-07-17: disclosed: GitHub Security Advisory GHSA-jcxw-h8ph-pxpv published
  • 2026-08-14: advisory: CVE-2026-73678 published on NVD

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.