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.
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.
Amazon's threat intelligence team used decoy servers to record AI agents scanning internet-exposed AI applications such as LiteLLM, One API, and Dify to steal embedded AI provider API keys. The captured records show stolen keys were then used to run the models that continued the hunt for more keys.
A report describes semantic caching vulnerabilities in LiteLLM, an LLM proxy and gateway. The source is a single Reddit security post with limited technical detail.
LiteLLM's MCP Streamable HTTP endpoint, prior to 1.84.0, let an unauthenticated attacker send a fabricated Authorization header to trigger an OAuth2 passthrough fallback. That fallback replaced failed LiteLLM key validation with an empty UserAPIKeyAuth() object, letting requests reach MCP tooling without a valid key. Affected: BerriAI LiteLLM before 1.84.0.