vLLM denial of service via prompt embeds on M-RoPE models

# vLLM denial of service via prompt embeds on M-RoPE models

https://gcdradar.vercel.app/incidents/vllm-denial-of-service-via-prompt-embeds-on-m-rope-models · AIR-2026-0813 · closed 2026-07-06

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability)
- Fix status: Fix status unknown
- First reported: 2026-07-06 (happened 2026-07-20)
- AI role: Attack on AI
- Products: vLLM
- CVE ids: CVE-2026-55514
- CVSS: 7.1

## Summary

Sending a pure prompt embeds payload in completion requests to vLLM models using M-RoPE causes an assertion failure. This flaw triggers a fatal crash that shuts down the entire inference server application.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-07-20: CVE-2026-55514: vLLM denial of service via prompt embeds on M-RoPE models (https://github.com/advisories/GHSA-33cg-gxv8-3p8g)
- 2026-07-06: CVE-2026-55514: vLLM is a library for LLM inference and serving. (https://nvd.nist.gov/vuln/detail/CVE-2026-55514)

## Sources

- 2026-07-06 NVD: CVE-2026-55514: vLLM is a library for LLM inference and serving. https://nvd.nist.gov/vuln/detail/CVE-2026-55514
- 2026-07-20 GitHub Advisory Database: CVE-2026-55514: vLLM denial of service via prompt embeds on M-RoPE models https://github.com/advisories/GHSA-33cg-gxv8-3p8g

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