Summary An assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLLM r…
| CVE ID | CVE-2026-41523 |
| Vendor | pip |
| Affected Product | vllm |
| Vulnerability Type | Vulnerability |
| CVSS Score | 7.5 (HIGH) |
| Actively Exploited | ❌ No known exploitation |
| Patch Status | See Vendor Advisory → |
| Reported By | CYBERDUDEBIVASH SENTINEL APEX Intelligence (via github_advisories) |
#
assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLLM runs in Python optimized mode (python -O or PYTHONOPTIMIZE=1).assert statement at [vllm/model_executor/layers/pooler/activations.py:48](https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/layers/pooler/activations.py#L48) as its sole security control to restrict which activation functions can be loaded from a HuggingFace model's config.json: ```pythonfunction_name: str | None = None if ( hasattr(config, "sentence_transformers") and "activation_fn" in config.sente
Sigma rules, YARA signatures, IOC table, and SIEM queries for Splunk, Elastic, Sentinel, and Chronicle — deployable in 5 minutes.