Description
Detects content exploiting CVE-2025-8083, unsafe deserialization in the torch_musa machine learning library that can allow code execution. The rule matches the pickled payloads that public exploits use to run code when the library loads model files.