Formal logic tool catches real AI systems failing fairness tests

Researchers developed a mathematical framework using formal logic to automatically verify whether AI systems meet ethical standards like fairness and explainability. When they tested it on two real-world AI systems—one used in criminal sentencing, one for loan decisions—the tool found both systems failed to meet basic fairness requirements, providing concrete evidence of discrimination that a machine could verify without human interpretation.

This shows a shift from AI ethics being debated in abstract terms to being measurable and verifiable using automated tools, which means problems become harder to ignore or defend.