Disclosed weakness · no CVE recorded

LiteLLM userless keys could read other teams’ key metadata

LiteLLM · LiteLLM Proxy · key information

First documented disclosure · Review checkpoint

Summary

A service-account key without a user could read other userless keys’ metadata across teams by key hash or alias.

Severity and exploitation evidence

No CVSS score recorded in this review.

Exploitation not established: The reviewed sources establish the weakness or a controlled demonstration; malicious exploitation against users is not established.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The owner comparison treated absent or empty user IDs as a shared owner.

Conditions required

The caller holds a userless virtual key and knows a target key hash or can guess its alias.

Potential impact

Disclosure of another key’s alias, team, permitted models, budgets, spend and metadata; the report does not establish plaintext-key disclosure.

Affected versions and configurations

The complete earlier affected-version range is unspecified.

Fix and mitigation status

Maintainer release notes identify 1.106.12, 1.108.13, 1.109.3 and 1.110.2; tagged 1.108.13 and 1.110.2 independently contain the non-empty owner check.

Practical next steps

Deploy the branch-specific correction and review cross-team access to key metadata.

Operating lesson

Missing ownership is not shared ownership.

Evidence and limits

The public PR predates the checkpoint on November 19; the stable releases are November 20. The follow-up restores a key’s own plaintext lookup without reopening cross-key reads. The maintainer’s later reproduction explicitly used one worker because reload ignored the worker flag; these are controlled checks, not an attack report.

Sources