Arbitrary Code Injection Affecting litellm package, versions [,1.85.0)
Threat Intelligence
The probability is the direct output of the EPSS model, and conveys an overall sense of the threat of exploitation in the wild. The percentile measures the EPSS probability relative to all known EPSS scores. Note: This data is updated daily, relying on the latest available EPSS model version. Check out the EPSS documentation for more details.
Do your applications use this vulnerable package?
In a few clicks we can analyze your entire application and see what components are vulnerable in your application, and suggest you quick fixes.
Test your applicationsSnyk Learn
Learn about Arbitrary Code Injection vulnerabilities in an interactive lesson.
Start learning- Snyk IDSNYK-PYTHON-LITELLM-19477048
- published1 Sept 2026
- disclosed31 Aug 2026
- creditYeran Gamage
Introduced: 31 Aug 2026
CVE-2026-37004 (opens in a new tab)How to fix?
Upgrade litellm to version 1.85.0 or higher.
Overview
litellm is a Library to easily interface with LLM API providers
Affected versions of this package are vulnerable to Arbitrary Code Injection in the /prompts/test endpoint through the rendered_content path in litellm/proxy/prompts/prompt_endpoints.py. An attacker can execute arbitrary OS commands by sending a crafted dotprompt_content parameter rendered in an unsandboxed Jinja2 environment. This lets an unauthenticated remote requester turn template expression traversal into code execution on the proxy host, putting the LiteLLM server and any data it handles at risk.
Notes:
- According to de28a4f, while
DotpromptManagerwas hardened to render throughImmutableSandboxedEnvironmentin 1.82.5, the three sibling managers (gitlab, arize, bitbucket) were missed and still instantiate plainjinja2.Environment()