BerriAI LiteLLM Unprotected Alternate Channel Vulnerability
Exploited in the wild, in CISA KEV since 18 Nov 2026 (due 2 Dec 2026), with 3 public exploits. Fixed in 1.83.10, 2.6-9 and 3.3.4.
LiteLLM through 2026-04-08 allows remote attackers to execute arbitrary code via bytecode rewriting at the /guardrails/test_custom_code URI.
Briefing
Written from the sources below · deepseek-v4.1-flash
LiteLLM through 2026-04-08 contains an unprotected alternate channel allowing remote attackers to execute arbitrary code via bytecode rewriting at the /guardrails/test_custom_code URI. BerriAI LiteLLM, litellm, and listed Red Hat products are affected.
Why it mattersExploitation is confirmed and the flaw is in CISA KEV with public exploit material and elevated EPSS, so exposed LiteLLM endpoints are at high risk of remote code execution.
ActionApply the vendor fix for LiteLLM and restrict access to /guardrails/test_custom_code.
Affected
Vendors, products and packages with version ranges
| Vendor | Product | Versions | Fixed |
|---|---|---|---|
| BerriAI | LiteLLM | =bb0639701796218a3447160e55c0f1097446e4e6085df7dfd39f476d4143743f | |
| LiteLLM | litellm | >=1.81.8 <1.83.10 | 1.83.10 |
| redhat | ansible_automation_platform | >=2.0 <2.6-9 | 2.6-9 |
| redhat | openshift_ai | >=3.3 <3.3.4 | 3.3.4 |
| redhat | openshift_lightspeed |
Exploit evidence
3 public sources
Coverage
News and research mentioning this CVE
Timeline
Every event on this vulnerability
Related
Other vulnerabilities in LiteLLM, strongest first
Embed
A live severity badge for an advisory, README or ticket