CVE-2023-37274: Auto-GPT path traversal in execute_python_code via basename argument

Summary

CVE-2023-37274 affects Auto-GPT versions before v0.4.3, where the `execute_python_code` command (introduced in v0.4.1) does not sanitize the `basename` argument. An LLM-supplied name such as `../../../main.py` lets code write outside the workspace directory, overwriting files like `autogpt/main.py`, which then runs on the host outside the Docker sandbox at the next start.

Mitigation

The issue has been patched in version 0.4.3. As a workaround, run Auto-GPT in a virtual machine or another environment where damage to files or corruption of the program is not a critical problem.

Related items

• MediumCVE-2026-121936: easyVoice server-side request forgery through openaiBaseUrl in /api/v1/ttsSimilar attack · NVD/CVE Database
• MediumCVE-2026-118628: MLflow get-artifacts server-side request forgery information disclosureSimilar attack · NVD/CVE Database
• HighCVE-2026-118625: MLflow Tracking Server privilege escalation via get_endpointsSimilar attack · NVD/CVE Database
• InfoWhen trust becomes the payload: Inside the threats, etc quarterly reportSimilar attack · Elastic Security Labs
• MediumCritical Langflow flaws enable authentication bypass and RCESimilar attack · CSO Online