CVE-2026-93841
vLLM through 0.29.0 contains a memory corruption vulnerability in the Triton _bincount_kernel where prompt token IDs index the penalty prompt-presence bitset without bounds checking against vocabulary size. Attackers can submit multimodal audio requests with tokens equal to vocabulary size, causing out-of-bounds writes that corrupt concurrent requests' sampler state and alter repetition penalty behavior.
- Published Sep 18, 2026
- CVSS 6.3 medium
- 0.2% chance of exploitation in the next 30 days (EPSS)
- A fix is available