---
id: CVE-2026-73558
title: vLLM is an inference and serving engine for large language models
summary: >-
  vLLM is an inference and serving engine for large language models. Prior to
  0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can
  cause act_and_mul_kernel to consume another batched user's input, allowing a
  reque…
severity: medium
cvss: 5.3
cvssVector: 'CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:H/I:N/A:N'
cwe:
  - CWE-190
vendor: vllm
product: vllm
affected:
  - vllm < 0.27.0
patched:
  - vllm 0.27.0
published: '2026-08-13'
updated: '2026-09-09'
sourceUpdated: '2026-09-09T20:58:37.713'
source: NVD
sourceUrl: 'https://nvd.nist.gov/vuln/detail/CVE-2026-73558'
references:
  - url: >-
      https://github.com/vllm-project/vllm/commit/451227cb3ff07989698fed982c2d3e4300257924
    label: security-advisories@github.com
  - url: 'https://github.com/vllm-project/vllm/issues/42860'
    label: security-advisories@github.com
  - url: 'https://github.com/vllm-project/vllm/pull/49660'
    label: security-advisories@github.com
  - url: 'https://github.com/vllm-project/vllm/releases/tag/v0.27.0'
    label: security-advisories@github.com
  - url: >-
      https://github.com/vllm-project/vllm/security/advisories/GHSA-7m6h-x95x-82q5
    label: security-advisories@github.com
  - url: >-
      https://github.com/vllm-project/vllm/security/advisories/GHSA-7m6h-x95x-82q5
    label: 134c704f-9b21-4f2e-91b3-4a467353bcc0
  - url: 'https://nvd.nist.gov/vuln/detail/CVE-2026-73558'
  - url: 'https://github.com/vllm-project/vllm'
  - url: 'https://github.com/advisories/GHSA-7m6h-x95x-82q5'
  - url: 'https://pypi.org/project/vllm'
tags:
  - nvd
  - osv
  - pip
  - ghsa
epss: 0.00404
epssPercentile: 0.31791
aliases:
  - GHSA-7m6h-x95x-82q5
  - PYSEC-2026-3935
ecosystem: pip
ingestedAt: '2026-09-08T21:11:12.323Z'
---

## Overview

vLLM is an inference and serving engine for large language models. Prior to 0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can cause act_and_mul_kernel to consume another batched user's input, allowing a request processed in the same inference batch to receive a partial or complete copy of another user's inference result. This issue is fixed in version 0.27.0.

## Remediation

Refer to the linked advisories for vendor-supplied fixes and affected version ranges.

## Package advisory (CVE-2026-73558)

Affected packages:

- `vllm < 0.27.0`

Patched in:

- `vllm 0.27.0`

Source: https://osv.dev/vulnerability/GHSA-7m6h-x95x-82q5