CVE-2026-115918: vLLM chat template parsing resource exhaustion

# CVE-2026-115918: vLLM chat template parsing resource exhaustion

https://gcdradar.vercel.app/incidents/cve-2026-115918-vllm-chat-template-parsing-resource-exhausti · AIR-2026-1542 · closed 2026-10-26

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability): Disclosed vulnerability with a fix and no exploitation reported; CVSS 6.5.
- Fix status: Patched
- First reported: 2026-10-26
- AI role: Attack on AI
- Products: vLLM
- CVE ids: CVE-2026-115918
- CVSS: 6.5

## Summary

A trust policy bypass in vLLM allows authenticated callers to supply custom chat templates that cause CPU or memory exhaustion during rendering. A patch has been released in version 0.32.0.

## What to do

Update to vLLM version 0.32.0.

## Affected and fixed versions

- Affected: vLLM 0.16.0 to 0.32.0
- Fixed in: vLLM 0.32.0

## Latest updates

- 2026-10-28: GitHub advisory published (https://github.com/advisories/GHSA-x567-49g5-p4cf)
- 2026-10-26: CVE-2026-115918: vLLM is an inference and serving engine for large language models. (https://nvd.nist.gov/vuln/detail/CVE-2026-115918)

## Sources

- 2026-10-26 NVD: CVE-2026-115918: vLLM is an inference and serving engine for large language models. https://nvd.nist.gov/vuln/detail/CVE-2026-115918
- 2026-10-28 GitHub Advisory Database: CVE-2026-115918: vLLM: Responses API bypasses trust_request_chat_template for chat_template_kwargs https://github.com/advisories/GHSA-x567-49g5-p4cf

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