Track vulnerabilities like this one.
Sign up free to get alerted when software you use is affected.
7.5
CVE-2023-37274: Auto-GPT allows attackers to run malicious Python code on the host system
CVE-2023-37274 · published 3 years ago
Summary
A bug in Auto-GPT's Python code execution feature allows an attacker to run malicious code on the host system if they have access to the Auto-GPT workspace directory. This can happen if Auto-GPT is run directly on the host system, not in a sandboxed environment. To protect against this, update to the latest version of Auto-GPT (0.4.3 or later) or run Auto-GPT in a virtual machine where damage to files is not a concern.
What to do
- Update agpt autogpt_classic to version 0.4.3 or later.
Affected software
| Vendor | Product | Affected versions |
|---|---|---|
| agpt | autogpt_classic |
< 0.4.3 cpe:2.3:a:agpt:autogpt_classic:*:*:*:*:*:*:*:* |
Original advisory text
Python code execution sandbox escape in non-docker version in Auto-GPT
Auto-GPT is an experimental open-source application showcasing the capabilities of the GPT-4 language model. When Auto-GPT is executed directly on the host system via the provided run.sh or run.bat files, custom Python code execution is sandboxed using a temporary dedicated docker container which should not have access to any files outside of the Auto-GPT workspace directory.
Before v0.4.3, the `execute_python_code` command (introduced in v0.4.1) does not sanitize the `basename` arg before writing LLM-supplied code to a file with an LLM-supplied name. This allows for a path traversal attack that can overwrite any .py file outside the workspace directory by specifying a `basename` such as `../../../main.py`. This can further be abused to achieve arbitrary code execution on the host running Auto-GPT by e.g. overwriting autogpt/main.py which will be executed outside of the docker environment meant to sandbox custom python code execution the next time Auto-GPT is started. The issue has been patched in version 0.4.3. As a workaround, the risk introduced by this vulnerability can be remediated by running Auto-GPT in a virtual machine, or another environment in which damage to files or corruption of the program is not a critical problem.
Before v0.4.3, the `execute_python_code` command (introduced in v0.4.1) does not sanitize the `basename` arg before writing LLM-supplied code to a file with an LLM-supplied name. This allows for a path traversal attack that can overwrite any .py file outside the workspace directory by specifying a `basename` such as `../../../main.py`. This can further be abused to achieve arbitrary code execution on the host running Auto-GPT by e.g. overwriting autogpt/main.py which will be executed outside of the docker environment meant to sandbox custom python code execution the next time Auto-GPT is started. The issue has been patched in version 0.4.3. As a workaround, the risk introduced by this vulnerability can be remediated by running Auto-GPT in a virtual machine, or another environment in which damage to files or corruption of the program is not a critical problem.
References
- https://github.com/CVEProject/cvelistV5/tree/main/cves/2023/37xxx/CVE-2023-37274... Vendor Advisory
- https://nvd.nist.gov/vuln/detail/CVE-2023-37274 Vendor Advisory
- https://github.com/Significant-Gravitas/Auto-GPT/pull/4756 Patch
- https://github.com/Significant-Gravitas/Auto-GPT/security/advisories/GHSA-5h38-m... Vendor Advisory
Internet-facing
14 days
Internal
At next upgrade
- Not known to be exploited
- Needs hands-on effort to exploit
- Gives an attacker full control
Type
CWE-94Code Injection
Timeline
Published13 Jul 2023
Updated4 Nov 2026
First seen2 Apr 2026
Track software like this
Free during beta