AI and data stack advisories

Severe, 6 weeks2431Projects328

2431 severe, 6 weeks · 328 projects

vLLMPYSEC-2026-3998

vLLM before 0.29.0 validates allowed_token_ids against tokenizer length instead...

vLLM

CVE-2026-93840 · Published Sep 18, 2026 · updated Sep 29, 2026

Medium5.3
Fix: upgrade to 0.29.0 or later
Source advisory

vLLM before 0.29.0 validates allowed_token_ids against tokenizer length instead of model output logits width in SamplingParams._validate_allowed_token_ids(). Attackers can supply token IDs above the output vocabulary that pass validation, causing LogitBiasState to corrupt GPU logits state and allow concurrent requests to sample tokens outside their allowlists.

Affected versions

PackageAffectedFixed in
vllm
PyPI
< 0.29.00.29.0
Details and references

More vLLM advisories

All vLLM