NewApp Monitoring is liveSign up free
AgentbeamBeta
AI agent vulnerability · LiteLLM

GHSA-g5ff-637f-6q2m: LiteLLM: internal_user_viewer arbitrary local file read via `vertex_ai_credentials`

highCVSS 3.1 8.1

A low-privilege LiteLLM proxy user can read any file the proxy process can read. They send a crafted `vertex_ai_credentials` value to `/utils/transform_request`, and the file contents go to a URL the attacker controls. All versions before 1.95.0 are affected. Upgrade to 1.95.0.

Published Oct 27, 2026 · Analysis Oct 27, 2026

What an attacker can do

LiteLLM's request-body safety blocklist rejects the vertex_credentials key. It missed the alias vertex_ai_credentials. The aliased value therefore reached the Vertex AI credential loader, which treats the value as a file path and reads it. The loaded content is then sent to an attacker-controlled URL (CWE-22, CWE-20).

The attacker needs only an authenticated account with the internal_user_viewer role. The advisory notes that role is obtainable through SSO login on many deployments, so the barrier is low where SSO is open to a broad domain.

The practical target is the proxy's own environment. Reading /proc/self/environ can reveal LITELLM_MASTER_KEY and the provider API keys. With the master key the attacker becomes proxy admin. Any other file the process can read, such as mounted config, service account JSON or Kubernetes tokens, is also exposed. The scope is confidentiality and integrity, with no availability impact in the CVSS vector.

Who is affected

All litellm versions below 1.95.0 are affected. The proxy server must be running, because the flaw is in the /utils/transform_request endpoint. A library-only use of LiteLLM without the proxy does not expose that endpoint.

Check the version:

pip show litellm

For container deployments, check the image tag of the proxy.

litellm

Ecosystem
PyPI
Affected
< 1.95.0
Fixed in
1.95.0
Check
pip show litellm

How to fix it

Upgrade to litellm 1.95.0 or later:

pip install --upgrade "litellm[proxy]>=1.95.0"

The fix adds vertex_ai_credentials to the request-body blocklist. No configuration change is needed. The advisory says there is no reliable configuration workaround on affected versions.

If you think the endpoint was abused, rotate LITELLM_MASTER_KEY and every provider API key the proxy holds.

How to tell if you were hit

Search proxy access logs for calls to /utils/transform_request by internal_user_viewer accounts, particularly request bodies that include vertex_ai_credentials with a filesystem path such as /proc/self/environ.

Check egress logs for outbound requests from the proxy to unfamiliar hosts around those calls. Look also for admin actions taken with the master key from unexpected addresses, such as new keys, teams or model entries.

Original advisory

From the GitHub Advisory Database (CC BY 4.0).

Summary

An authenticated LiteLLM proxy user could read arbitrary local files from the proxy runtime by sending a crafted vertex_ai_credentials value to /utils/transform_request. The request-body safety blocklist rejected vertex_credentials but missed its alias vertex_ai_credentials, which let the value reach the Vertex AI credential loader and trigger a file read that is sent to an attacker-controlled URL.

Impact

A low-privilege authenticated user (only internal_user_viewer, obtainable through SSO login on many deployments) could read any file readable by the LiteLLM process. Reading files such as /proc/self/environ can disclose LITELLM_MASTER_KEY and provider API keys, so the impact can escalate to full proxy-admin control.

Remediation

Upgrade to 1.95.0 or later. The fix adds the missing alias to the request-body blocklist, so no configuration change is needed once you upgrade. There is no reliable configuration workaround on affected versions.

Credited to rekter0.

Questions

Which LiteLLM role can exploit this file read?

internal_user_viewer, the lowest internal role. The advisory says it is obtainable through SSO login on many deployments, so no admin access is needed.

Can this lead to full proxy takeover?

Yes. Reading /proc/self/environ can disclose LITELLM_MASTER_KEY and provider API keys, which the advisory says can escalate to full proxy-admin control.

Is there a workaround short of upgrading?

No. The advisory says there is no reliable configuration workaround on affected versions. Upgrade to 1.95.0 or later.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology

Coverage on the Beam blog

More LiteLLM vulnerabilities

All LiteLLM entries