CVE-2026-44222: vLLM vulnerable to remote DoS via special-token placeholders
vLLM
Unauthenticated prompts spelling special tokens caused vLLM to index into empty grids during input-position computation. This triggered an unhandled IndexError terminating the worker and causing availability degradation.
disclosed CVENo confirmed exploitation, so this record is tracked as a disclosed vulnerability and not counted as an incident. It moves into the counts if CISA or the vendor confirms exploitation. Why
- CVSS
- 6.5
- EPSS
- <0.01 · top 64% · 2026-11-13
- fix status
- Fix status unknown
- first report
- 5 May 2026
- last update
- 12 May 2026
- CVE ids
Timeline · 2
Sources · 2
Advisories and databases
- 2026-05-05CVE-2026-44222: vLLM Vulnerable to Remote DoS via Special-Token PlaceholdersGitHub Advisory Database · github.com/advisories/GHSA-hpv8-x276-m59f
- 2026-05-12CVE-2026-44222: vLLM is an inference and serving engine for large language models (LLMs).NVD · nvd.nist.gov/vuln/detail/CVE-2026-44222
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Patch now
- AI role
- Attack on AI
- categories
- Infrastructure vulnerability
- severity
- 2 · Disclosed vulnerability
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply Chain
- mitre atlas
- no technique for these categories
- dates
- happened 5 May 2026 · first reported 5 May 2026
- origin
- GitHub Advisory DatabaseVulnerability record from NVD / the GitHub Advisory Database. This product uses the NVD API but is not endorsed or certified by the NVD.
- record
- confidence 0.95 · created 2026-10-18 07:00Z · updated 2026-10-18 10:04Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# CVE-2026-44222: vLLM vulnerable to remote DoS via special-token placeholders https://gcdradar.vercel.app/incidents/cve-2026-44222-vllm-vulnerable-to-remote-dos-via-special-tok · AIR-2026-0616 · closed 2026-05-05 - Lane: Patch now - Severity: 2 (Disclosed vulnerability) - Fix status: Fix status unknown - First reported: 2026-05-05 - AI role: Attack on AI - Products: vLLM - CVE ids: CVE-2026-44222 - CVSS: 6.5 ## Summary Unauthenticated prompts spelling special tokens caused vLLM to index into empty grids during input-position computation. This triggered an unhandled IndexError terminating the worker and causing availability degradation. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2026-05-12: CVE-2026-44222: vLLM is an inference and serving engine for large language models (LLMs). (https://nvd.nist.gov/vuln/detail/CVE-2026-44222) - 2026-05-05: CVE-2026-44222: vLLM Vulnerable to Remote DoS via Special-Token Placeholders (https://github.com/advisories/GHSA-hpv8-x276-m59f) ## Sources - 2026-05-05 GitHub Advisory Database: CVE-2026-44222: vLLM Vulnerable to Remote DoS via Special-Token Placeholders https://github.com/advisories/GHSA-hpv8-x276-m59f - 2026-05-12 NVD: CVE-2026-44222: vLLM is an inference and serving engine for large language models (LLMs). https://nvd.nist.gov/vuln/detail/CVE-2026-44222 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closed6.5vLLM vLLM speculative decoding crashes server via penalty parameters
Patch now · Disclosed · since 6 May 2026 · updated 12 May 2026 · CVE-2026-44223
AI - closed5.6vLLM CVE-2026-7141: vLLM Uninitialized Resource Vulnerability
Patch now · Disclosed · since 27 Apr 2026 · CVE-2026-7141
AI - closed7.5vLLM CVE-2026-115912: vLLM shared-memory multimodal remote code execution
Patch now · Disclosed · Patched · since 26 Oct 2026 · updated 28 Oct 2026 · CVE-2026-115912
AI - closed6.5vLLM CVE-2026-115916: vLLM tokenizer off-by-one engine termination
Patch now · Disclosed · Patched · since 26 Oct 2026 · updated 28 Oct 2026 · CVE-2026-115916
AI - closed6.5vLLM CVE-2026-115918: vLLM chat template parsing resource exhaustion
Patch now · Disclosed · Patched · since 26 Oct 2026 · updated 28 Oct 2026 · CVE-2026-115918
AI - closed6.5vLLM CVE-2026-115914: vLLM multimodal OCR path denial of service
Patch now · Disclosed · Patched · since 26 Oct 2026 · updated 28 Oct 2026 · CVE-2026-115914
AI