{
 "id": "AIR-2026-0766",
 "slug": "cve-2026-54235-vllm-temperature-validation-bypass-for-nan-an",
 "title": "CVE-2026-54235: vLLM temperature validation bypass for NaN and Infinity values",
 "summary": "vLLM fails to properly validate NaN and positive Infinity temperature parameters due to standard float comparison limitations, propagating invalid values to GPU kernels and causing crashes. The issue is documented as a vendor advisory.",
 "type": "attack_on_ai",
 "categories": [
  "infra_vuln"
 ],
 "severity": 2,
 "targets": [
  "framework_or_tooling"
 ],
 "products": [
  "vLLM"
 ],
 "incident_date": "2026-06-17",
 "incident_date_precision": "day",
 "first_reported": "2026-06-17",
 "sources": [
  {
   "url": "https://github.com/advisories/GHSA-7h4p-rffg-7823",
   "publisher": "GitHub Advisory Database",
   "title": "CVE-2026-54235: vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels",
   "published_at": "2026-06-17"
  },
  {
   "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-54235",
   "publisher": "NVD",
   "title": "CVE-2026-54235: vLLM is an inference and serving engine for large language models (LLMs).",
   "published_at": "2026-06-22"
  }
 ],
 "source_count": 2,
 "confidence": 0.95,
 "status": "published",
 "origin": "seed_ghsa",
 "tags": [
  "CVE-2026-54235",
  "vllm",
  "denial-of-service"
 ],
 "created_at": "2026-10-18T07:00:14.101Z",
 "updated_at": "2026-10-18T10:04:21.885Z"
}