AI Regulatory & FDA

The European Union recently passed the AI Act to regulate artificial intelligence, especially in high-risk areas. However, the law does not clearly define what makes a system 'AI' versus a simple calculator. It hinges on whether a system can 'infer' or make independent deductions from data. Researchers created a new framework to measure this capability. By testing it on credit scoring systems, they discovered that we must look at the entire data journey, including human involvement, to decide if a system qualifies as AI. For regular people, this research is a vital first step toward ensuring the algorithms used in healthcare and finance are properly regulated and safe.

The European Union recently passed the AI Act to regulate artificial intelligence, especially in high-risk areas. However, the law only applies to systems that have the 'capability to infer'—meaning they can make decisions or predictions on their own—and it does not clearly define what this means. Researchers created a new framework to measure different levels of this decision-making capability. By testing it on credit scoring systems, they found that we must look at the entire data process, including human involvement, to decide if a system counts as AI. This matters because it determines which computer programs must follow strict safety rules before being used on the public.