Bristol researchers propose drug approval model for AI safety

VERA Brief

AI-generated. Grounded in the article and its cited sources.

Researchers at the University of Bristol have proposed a new framework for evaluating the safety of artificial intelligence systems in medicine. This framework adapts existing drug approval processes to assess system limitations, fairness, and clinical suitability, aiming to ensure AI tools are safe and equitable for all patients.

Key facts

• Researchers at the University of Bristol have proposed a framework for evaluating the safety of medical artificial intelligence systems.
• The proposed framework draws parallels with existing drug approval processes in medicine.
• The framework focuses on assessing system limitations, fairness across diverse patient demographics, and clinical suitability.
• The initiative aims to identify AI models that may pose risks in real-world clinical applications despite technical adequacy.
• The goal is to ensure AI tools used in healthcare are reliably safe and equitable for all patients.