CVE-2026-93989
vLLM through 0.29.0 Cross-Request Logits Corruption via bad_words
CVSS 4.0 2.3 LOWCVSS 3.1 3.1 LOWCheck source for fixed versionsKEV status unknown
cve-program · CVE-2026-93989
PublishedvLLM through 0.29.0 fails to properly validate bad_words token indices against the model's generation output width in SamplingParams.update_from_tokenizer(). Attackers can supply out-of-bounds token indices that corrupt logits memory of concurrent requests, causing different in-flight HTTP requests to return incorrect tokens.
Read the publisher's advisoryAffected versions, risk and source history
Related advisories and sources
Publishers link these advisories through the identifiers shown below. A shared identifier or date alone does not tell you whether your software is affected or whether the advisories describe the same cause.
CVE-2026-93989 · CVE overview and exploitation status
cve-program · CVE-2026-93989
- https://github.com/vllm-project/vllm/pull/48824
- https://github.com/vllm-project/vllm/blob/98dff2a81d747d1dba01a47f939f48c3526d4206/vllm/v1/worker/gpu/sample/bad_words.py
- https://github.com/vllm-project/vllm/blob/98dff2a81d747d1dba01a47f939f48c3526d4206/vllm/sampling_params.py#L694-L753
- https://github.com/vllm-project/vllm
- https://www.vulncheck.com/advisories/vllm-through-0.29.0-cross-request-logits-corruption-via-bad-words
Related changes
Linked by CVE identifierNo related changes are available here.