Unsloth Studio Flaw Turns Routine Model Inspection Into Code Execution

Patched Unsloth Studio flaw let malicious AI models run arbitrary Python code on researchers' machines via trust_remote_code during model inspection.

A vulnerability in Unsloth Studio allowed a malicious AI model to achieve arbitrary Python code execution when a user inspected it, abusing the trust_remote_code setting. The flaw has been patched. Unsloth Studio is a popular toolkit for fine-tuning and working with open-weight LLMs, so users who inspect untrusted models were at risk of code execution on their local machines.

• Malicious models could execute arbitrary Python code during inspection
• Abuses the trust_remote_code setting in Unsloth Studio
• Flaw is already patched; no active exploitation reported
• Highlights risk of inspecting untrusted open-weight models