vLLM: Cross-User Data Leak Vulnerability

Answer in brief

CVE-2026-73558 records a Medium severity (CVSS 5.3) vulnerability in vLLM: Cross-User Data Leak Vulnerability. The current sources do not mark it as known exploited. The current feed maps vllm-project/vllm (generic), vllm (pip), vllm (pypi). Check affected ranges and fixed versions before updating.

Analysis pending evidence review

HOL Guard separates source facts from reviewed analysis. See the methodology.

Published Aug 13, 2026Updated Sep 9, 2026Source checked Sep 27, 2026First seen by HOL Aug 13, 2026Material review Sep 9, 2026

Upstream Advisory

Key facts

Risk: Medium · CVSS 5.3
Exploitation: Not marked as known exploited
Affected software: 3 mapped packages or products
Fix availability: Available

Why this deserves its current priority

CVSS is 5.3. The current sources do not mark it as known exploited. Treat this as a source-backed prioritization signal, not a statement about your environment.

Analysis status

Analysis pending evidence review

Factual feed record only; HOL analysis is not approved for indexing. Read the methodology.

Affected scope and exposure questions

The current feed maps vllm-project/vllm (generic), vllm (pip), vllm (pypi). Check affected ranges and fixed versions before updating.

Mapped affected packages and fixed versions
Package	Affected range	Fixed version
vllm-project/vllmgeneric	< 0.27.0	Not reported
vllmpip	<0.27.0	0.27.0
vllmpypi	>=0 <0.27.0	0.27.0

Recommended response

Check inventory. Check lockfiles and deployed manifests for vllm-project/vllm, vllm, vllm.
Review the reported fix. Update vllm to 0.27.0; vllm to 0.27.0 if you use the affected versions. Test the change in a non-production environment first.

Evidence timeline and material changes

Published upstream
Aug 13, 2026
Evidence: source:cvelist:source_dates:source-dates:record
Source modified
Sep 9, 2026
Evidence: source:cvelist:source_dates:source-dates:record
First seen by HOL
Aug 13, 2026

Sources and claim methodology

Upstream source description

vLLM is an inference and serving engine for large language models. Prior to 0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can cause act_and_mul_kernel to consume another batched user's input, allowing a request processed in the same inference batch to receive a partial or complete copy of another user's inference result. This issue is fixed in version 0.27.0.

Quoted source text, attributed separately from HOL analysis.