This research systematically evaluates security risks when deploying large language models in clinical environments through comprehensive jailbreaking tests across seven LLMs.

• Proposes a domain-adapted evaluation pipeline specifically for medical contexts
• Identifies critical vulnerabilities where AI systems could be manipulated to provide harmful medical information
• Demonstrates the need for enhanced safeguards before deploying LLMs in sensitive healthcare applications

As AI increasingly enters clinical practice, this work highlights the urgent need to address security gaps that could compromise patient safety and medical ethics.

Towards Safe AI Clinicians: A Comprehensive Study on Large Language Model Jailbreaking in Healthcare