{"action":"Upgrade vllm to 0.30.0 or later.","affected":[{"fixed":"0.30.0","introduced":"0"}],"component":"vllm","cve_ids":["CVE-2026-93989"],"ecosystem":"PyPI","epss":0.00202,"exploited":false,"fixed_versions":["0.30.0"],"id":"tcs_01M5KH96M8HR2XHCABHH5RY1VC","modified_at":"2026-09-29T09:00:02.993282989Z","published_at":"2026-09-19T23:17:10.043Z","references":["https://osv.dev/vulnerability/PYSEC-2026-4000","https://github.com/vllm-project/vllm/blob/98dff2a81d747d1dba01a47f939f48c3526d4206/vllm/sampling_params.py#L694-L753","https://github.com/vllm-project/vllm/blob/98dff2a81d747d1dba01a47f939f48c3526d4206/vllm/v1/worker/gpu/sample/bad_words.py","https://www.vulncheck.com/advisories/vllm-through-0.29.0-cross-request-logits-corruption-via-bad-words","https://github.com/vllm-project/vllm/pull/48824","https://github.com/vllm-project/vllm"],"severity":{"label":"medium","score":4.3},"sources":[{"source":"osv","source_id":"PYSEC-2026-4000","url":"https://osv.dev/vulnerability/PYSEC-2026-4000"}],"summary":"vLLM 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 i"}