Unsloth Studio Flaw Turns Routine Model Inspection Into Code Execution

A vulnerability in Unsloth Studio has been patched that could allow malicious AI models to execute arbitrary Python code. This code execution occurs during the model inspection process when a specific setting, trust_remote_code, is enabled. The flaw enables malicious models to perform unauthorized actions. The vulnerability has been addressed with a patch. The issue is related to the trust_remote_code setting, which plays a role in the code execution process. The patched vulnerability prevents malicious AI models from exploiting this flaw during model inspection.