github.com
https://github.com/vllm-project/vllm/commit/793cf79c89d4049124e756915468ac30318f2e50 CVE-2026-73557
MEDIUM
vLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts
Record summary
CVE-2026-73557 has a selected CVSS score of 6.3 (medium).
Description
vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, and restore state can be raced by concurrent prompt_embeds parts submitted to POST /v1/chat/completions through AsyncMultiModalItemTracker.resolve_items, asyncio.gather, and the default executor, allowing an invalid sparse tensor to reach tensor.to_dense despite the CVE-2025-62164 guard when enable_prompt_embeds is enabled. This issue is fixed in version 0.26.0.
Description source: CVE List
Exploitation context
CISA SSVC decision
ExploitationNone
AutomatableNo
Technical impactPartial
CISA Coordinator · SSVC 2.0.3 · Evaluated Aug 13, 2026 · Source: CVE List
Affected products and versions
1| Product | Source | Version range | Status |
|---|---|---|---|
| CVE List | >= 0.20.2rc0, < 0.26.0 | affected |
References
4github.com
https://github.com/vllm-project/vllm/pull/48583 github.com
https://github.com/vllm-project/vllm/releases/tag/v0.26.0 github.comConfirmation
https://github.com/vllm-project/vllm/security/advisories/GHSA-pr7f-p5mw-fc87