CVE-2026-115911: vLLM chat template denial of service vulnerability

# CVE-2026-115911: vLLM chat template denial of service vulnerability

https://gcdradar.vercel.app/incidents/cve-2026-115911-vllm-chat-template-denial-of-service-vulnera · AIR-2026-1537 · closed 2026-10-26

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability): Disclosed vulnerability with a patch available and no confirmed exploitation.
- Fix status: Patched
- First reported: 2026-10-26
- AI role: Attack on AI
- Products: vLLM
- CVE ids: CVE-2026-115911
- CVSS: 6.5

## Summary

An unauthenticated remote user can supply a crafted chat template that causes resource exhaustion and blocks responses in vLLM. The vulnerability affects versions from 0.19.0 until 0.32.0.

## What to do

Update to vLLM 0.32.0 or later.

## Affected and fixed versions

- Affected: vLLM >= 0.19.0, < 0.32.0
- Fixed in: vLLM 0.32.0

## Latest updates

- 2026-10-28: GitHub advisory published (https://github.com/advisories/GHSA-542m-9234-8gvf)
- 2026-10-26: CVE-2026-115911: vLLM is an inference and serving engine for large language models. (https://nvd.nist.gov/vuln/detail/CVE-2026-115911)

## Sources

- 2026-10-26 NVD: CVE-2026-115911: vLLM is an inference and serving engine for large language models. https://nvd.nist.gov/vuln/detail/CVE-2026-115911
- 2026-10-28 GitHub Advisory Database: CVE-2026-115911: vLLM: DoS via unbounded chat_template Jinja2 rendering exhausts the renderer thread pool https://github.com/advisories/GHSA-542m-9234-8gvf

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.