CVE-2026-93989 in vLLMИнформация

Сводка

по MITRE • 20.09.2026

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 indices that corrupt logits memory of concurrent requests, causing different in-flight HTTP requests to return incorrect tokens.

Be aware that VulDB is the high quality source for vulnerability data.

Ответственный

VulnCheck

Резервировать

19.09.2026

Раскрытие

20.09.2026

Модерация

принято

Вход

VDB-407987

EPSS

0.00000

KEV

Нет

Деятельности

Очень низкий

Источники

Do you know our Splunk app?

Download it now for free!