CVE-2026-93989

vLLM through 0.29.0 is affected by a validation flaw in SamplingParams.update_from_tokenizer(). The function fails to properly validate bad_words token indices against the model's generation output width, allowing attackers to supply out-of-bounds indices that corrupt the logits memory of concurr...

CVE-2026-93840

vLLM before 0.29.0 is vulnerable to cross-request logits corruption via the allowed_token_ids parameter. The root cause is in SamplingParams._validate_allowed_token_ids() , which validates token IDs against the tokenizer length rather than the model output logits width. An attacker can supply tok...