Is CVE-2026-89032 exploited in the wild?

No. Junglewise Threat Intelligence has no report of CVE-2026-89032 being exploited in the wild.

How severe is CVE-2026-89032?

High severity, with a CVSS score of 7.7, published 25 September 2026.

Which products and versions are affected?

Berriai LiteLLM before 1.101.0-rc.1.

Is there a fix for CVE-2026-89032?

A patch was published on 2026-09-03: Fix merged in commit 16db51e; semantic_cache_scope parameter added to isolate cache by end-user Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

How is CVE-2026-89032 attacked?

It is attacked over the network.

Am I affected by CVE-2026-89032?

Check whether you run Berriai LiteLLM at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

CVE-2026-89032: BerriAI LiteLLM tenant isolation bypass in semantic cache

CVE-2026-89032 · Severity: high · CVSS 7.7 · Published 2026-09-25

BerriAI LiteLLM is an AI platform that caches LLM (large language model) responses to improve performance. In affected versions before 1.101.0-rc.1, authenticated users sharing a virtual key can read other users' cached responses, exposing sensitive data like personally identifiable information, financial records, and source code. Attackers can also trick the system into executing malicious tool calls under victim credentials.

Technologies: Berriai LiteLLM. Vendors: Berriai.

Quick answers

Is CVE-2026-89032 exploited in the wild?: No. Junglewise Threat Intelligence has no report of CVE-2026-89032 being exploited in the wild.
How severe is CVE-2026-89032?: High severity, with a CVSS score of 7.7, published 25 September 2026.
Which products and versions are affected?: Berriai LiteLLM before 1.101.0-rc.1.
Is there a fix for CVE-2026-89032?: A patch was published on 2026-09-03: Fix merged in commit 16db51e; semantic_cache_scope parameter added to isolate cache by end-user Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
How is CVE-2026-89032 attacked?: It is attacked over the network.
Am I affected by CVE-2026-89032?: Check whether you run Berriai LiteLLM at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

Technical details

The semantic cache layer fails to include tenant scope metadata (end_user identifier) when generating cache keys, causing a key mismatch between _get_semantic_cache_tenant_scope() and _get_metadata_variable_name(). Authenticated users with a valid virtual key can submit semantically similar prompts to routes like /v1/responses, /v1/chat/completions, and /bedrock/* to retrieve cached responses intended for other tenants, including function_call and tool_calls payloads. The vulnerability requires valid virtual key authentication but no additional user interaction.

Affected products

BerriAI LiteLLM before 1.101.0-rc.1

Timeline

2026-09-25: disclosed
2026-09-03: patched: Fix merged in commit 16db51e; semantic_cache_scope parameter added to isolate cache by end-user

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.