LiteLLM Landed on CISA's Exploited Vulnerabilities List Twice in a Month

LiteLLM, the open-source gateway that routes requests across LLM providers and sits inside a large share of production AI stacks, had two critical vulnerabilities added to CISA's Known Exploited Vulnerabilities catalog within a single month.

The first, CVE-2026-42208, is a SQL injection flaw with a CVSS score of 9.3, exploitable through a crafted Authorization header. CISA added it to the KEV catalog on May 8, 2026. The second, CVE-2026-42271, is a command injection vulnerability with a CVSS score of 8.7, affecting LiteLLM versions 1.74.2 through 1.83.6. CISA added it a month later, on June 8, 2026, after confirming active exploitation in the wild.

The Hacker News and CSA Labs both flagged the same underlying concern: two critical, actively exploited vulnerabilities in the same widely deployed gateway within four weeks is not a coincidence of bad luck, it's a sign that LiteLLM is being actively probed as a high-value target, precisely because so many AI applications route through it as a single chokepoint.

If you're running LiteLLM anywhere in your stack, the version range for the command injection flaw, 1.74.2 to 1.83.6, is worth checking immediately, and both CVEs being KEV-listed means CISA has already confirmed exploitation is not theoretical. Patch first, audit access logs for the Authorization header pattern second.