Code Injection Vulnerability in LiteLLM AI Gateway Affecting BerriAI
CVE-2026-59821

2.1LOW

Key Information:

Vendor

Berriai

Status
Litellm
Vendor
CVE Published:
8 July 2026

What is CVE-2026-59821?

LiteLLM, an AI Gateway used to interface with LLM APIs, contains a vulnerability in its Custom Code Guardrails feature prior to version 1.82.0-stable. This issue arises when a privileged user can create or update guardrails without the appropriate sandboxing and validation that is enforced on the test endpoint. As a result, the user could inject custom Python code that would run in the LiteLLM proxy environment, potentially exposing sensitive secrets that the process can access. This vulnerability has been addressed in version 1.82.0-stable.

Affected Version(s)

litellm < 1.82.0-stable

References

https://github.com/BerriAI/litellm/security/advisories/GH...
x_refsource_CONFIRM
https://github.com/BerriAI/litellm/commit/e50b4486d0f7aa0...
x_refsource_MISC
https://github.com/BerriAI/litellm/releases/tag/v1.82.0-s...
x_refsource_MISC

CVSS V4

Score:
2.1
Severity:
LOW
Confidentiality:
Low
Integrity:
Low
Availability:
Low
Attack Vector:
Network
Attack Complexity:
Low
Attack Required:
Physical
Privileges Required:
Undefined
User Interaction:
None

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.