gcdRadar
archive/Patch now/AIR-2026-0081
closedclosed 13 Aug 2026Patch now2 Disclosed vulnerabilityAI

CVE-2026-73558: vLLM data leakage vulnerability in activation kernels

vLLM

An integer overflow in vLLM activation kernels allowed inference requests in the same batch to consume another user's input, leaking inference results. The issue was patched in version 0.27.0.

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
5.3
EPSS
<0.01 · top 64% · 2026-11-13
EPSS scores by FIRST
fix status
Fix status unknown
first report
13 Aug 2026
last update
8 Sep 2026
CVE ids
CVE-2026-73558other records

Timeline · 2

  1. 13 Aug 2026CVE-2026-73558: vLLM is an inference and serving engine for large language models.
  2. 8 Sep 2026CVE-2026-73558: vLLM: Cross-User Data Leak Vulnerability

Sources · 2

Advisories and databases

  1. 2026-08-13
    CVE-2026-73558: vLLM is an inference and serving engine for large language models.
    NVD · nvd.nist.gov/vuln/detail/CVE-2026-73558
  2. 2026-09-08
    CVE-2026-73558: vLLM: Cross-User Data Leak Vulnerability
    GitHub Advisory Database · github.com/advisories/GHSA-7m6h-x95x-82q5

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 vulnerabilityData leak
severity
2 · Disclosed vulnerability
fix status
Fix status unknown
targets
Framework / tooling
owasp llm top 10
LLM02:2025 Sensitive Information DisclosureLLM03:2025 Supply Chain
mitre atlas
no technique for these categories
tags
vllm
dates
happened 13 Aug 2026 · first reported 13 Aug 2026
origin
NVD
Vulnerability 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 06:33Z · updated 2026-10-18 07:00Z

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-2026-73558: vLLM data leakage vulnerability in activation kernels

https://gcdradar.vercel.app/incidents/cve-2026-73558-vllm-data-leakage-vulnerability-in-activation · AIR-2026-0081 · closed 2026-08-13

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability)
- Fix status: Fix status unknown
- First reported: 2026-08-13
- AI role: Attack on AI
- Products: vLLM
- CVE ids: CVE-2026-73558
- CVSS: 5.3

## Summary

An integer overflow in vLLM activation kernels allowed inference requests in the same batch to consume another user's input, leaking inference results. The issue was patched in version 0.27.0.

## What to do

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

## Latest updates

- 2026-09-08: CVE-2026-73558: vLLM: Cross-User Data Leak Vulnerability (https://github.com/advisories/GHSA-7m6h-x95x-82q5)
- 2026-08-13: CVE-2026-73558: vLLM is an inference and serving engine for large language models. (https://nvd.nist.gov/vuln/detail/CVE-2026-73558)

## Sources

- 2026-08-13 NVD: CVE-2026-73558: vLLM is an inference and serving engine for large language models. https://nvd.nist.gov/vuln/detail/CVE-2026-73558
- 2026-09-08 GitHub Advisory Database: CVE-2026-73558: vLLM: Cross-User Data Leak Vulnerability https://github.com/advisories/GHSA-7m6h-x95x-82q5

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

Related

  1. closed5.3
    vLLM CVE-2026-73555: Information disclosure vulnerability in vLLM request validation handling

    Patch now · Disclosed · since 13 Aug 2026 · updated 4 Sep 2026 · CVE-2026-73555

    AI
  2. closed8.8
    vLLM CVE-2026-116176: vLLM server-side request forgery leads to NCCL unique_id and KV-cache leakage

    Patch now · Disclosed · since 25 Oct 2026 · CVE-2026-116176

    AI
  3. closed7.1
    vLLM CVE-2026-114054: vLLM cache aliasing vulnerability exposes processed media and KV state

    Patch now · Disclosed · since 21 Oct 2026 · CVE-2026-114054

    AI
  4. closed7.5
    vLLM CVE-2025-30202: Data Exposure via ZeroMQ in Multi-Node vLLM Deployment

    Patch now · Disclosed · since 29 Apr 2025 · updated 30 Apr 2025 · CVE-2025-30202

    AI
  5. closed5.0
    Langflow CVE-2026-79723: Information disclosure vulnerability in IBM Langflow OSS

    Patch now · Disclosed · since 10 Sep 2026 · CVE-2026-79723

    AI
  6. closed8.7
    Open WebUI CVE-2026-87995: Open WebUI XSS via terminal port-preview iframe sandbox misconfiguration

    Patch now · Disclosed · since 10 Sep 2026 · updated 10 Sep 2026 · CVE-2026-87995

    AI