CVE-2026-115912: vLLM shared-memory multimodal remote code execution

# CVE-2026-115912: vLLM shared-memory multimodal remote code execution

https://gcdradar.vercel.app/incidents/cve-2026-115912-vllm-shared-memory-multimodal-remote-code-ex · AIR-2026-1544 · closed 2026-10-26

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability): Disclosed high-severity vulnerability with a fix and no exploitation reported; CVSS 7.5.
- Fix status: Patched
- First reported: 2026-10-26
- AI role: Attack on AI
- Products: vLLM
- CVE ids: CVE-2026-115912
- CVSS: 7.5

## Summary

Improper trust handling of shared-memory multimodal features in vLLM scale-out deployments allows authenticated callers to trigger arbitrary Python execution via insecure deserialization. Version 0.32.0 addresses this vulnerability.

## What to do

Update to vLLM version 0.32.0.

## Affected and fixed versions

- Affected: vLLM < 0.32.0
- Fixed in: vLLM 0.32.0

## Latest updates

- 2026-10-28: GitHub advisory published (https://github.com/advisories/GHSA-6757-hggm-hp2w)
- 2026-10-26: CVE-2026-115912: vLLM is an inference and serving engine for large language models. (https://nvd.nist.gov/vuln/detail/CVE-2026-115912)

## Sources

- 2026-10-26 NVD: CVE-2026-115912: vLLM is an inference and serving engine for large language models. https://nvd.nist.gov/vuln/detail/CVE-2026-115912
- 2026-10-28 GitHub Advisory Database: CVE-2026-115912: vLLM: Public multimodal SHM handle forgery reaches worker-side `pickle.loads()` with insecure serialization disabled https://github.com/advisories/GHSA-6757-hggm-hp2w

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