CVE-2026-44223

[CVE catalog]

Severity

High · CVSS 7.5CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

Published

Apr 28, 2026, 10:00:00 PM UTC

Source

[NVD] · [CVE record]

EPSS

0.37% · Oct 25, 2026

A flaw was found in vLLM, an inference and serving engine for large language models (LLMs). The extract_hidden_states speculative decoding proposer returns a tensor with an incorrect shape after the first decode step. This can be triggered by a remote attacker sending a request that uses sampling penalty parameters, such as repetition_penalty, frequency_penalty, or presence_penalty. Successful exploitation leads to a RuntimeError that crashes the EngineCore process, resulting in a denial of service (DoS). vllm: vLLM: Denial of Service via malformed tensor shape in speculative decoding This Important denial of service flaw in vLLM, as used in Red Hat AI Inference Server and Red Hat OpenShift AI, allows a remote attacker to crash the EngineCore process. By sending a request with specific sampling penalty parameters, an attacker can trigger an incorrect tensor shape, leading to a service disruption for affected AI inference workloads. The CVSS score(s) listed for this vulnerability do not reflect the associated product's status, and are included for informational purposes to better understand the severity of this vulnerability.