---
id: CVE-2026-114049
title: vLLM before 0.31.0 Stale allowed_token_ids Mask via InputBatch.condense
summary: >-
  vLLM before 0.31.0 contains a cross-request mask leak in InputBatch.condense()
  that allows remote attackers to constrain other clients' outputs by leaving
  stale allowed_token_ids rows. Attackers can submit concurrent constrained
  generate…
severity: low
cvss: 3.7
cvssVector: 'CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N'
cvssSource: cna
cwe:
  - CWE-440
vendor: vllm-project
product: vllm
affected:
  - vllm < 0.31.0
published: '2026-10-21'
updated: '2026-10-21'
sourceUpdated: '2026-10-21T00:36:26.938Z'
source: CVEORG
sourceUrl: 'https://www.cve.org/CVERecord?id=CVE-2026-114049'
references:
  - url: >-
      https://github.com/vllm-project/vllm/security/advisories/GHSA-6cxc-2vcg-w5qc
    label: GitHub Security Advisory (GHSA-6cxc-2vcg-w5qc)
  - url: >-
      https://www.vulncheck.com/advisories/vllm-before-0.31.0-stale-allowed-token-ids-mask-via-inputbatch-condense
    label: >-
      VulnCheck Advisory: vLLM before 0.31.0 Stale allowed_token_ids Mask via
      InputBatch.condense
tags:
  - cve.org
ingestedAt: '2026-10-21T01:37:08.050Z'
---

## Overview

vLLM before 0.31.0 contains a cross-request mask leak in InputBatch.condense() that allows remote attackers to constrain other clients' outputs by leaving stale allowed_token_ids rows. Attackers can submit concurrent constrained generate requests, including via unauthenticated /invocations, so recycled batch slots apply their token allowlist to victims' unrestricted requests.

## Affected

- `vllm < 0.31.0`

## Remediation

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