From Protocols to Evidence: Bounded Claims for AI in Service of the Common Good
Moving beyond procedural compliance like the EU AI Act and NIST AI RMF, this paper examines how AI deployments interact with existing institutional shortcomings. The authors introduce a 'rupture test' to evaluate systems against institutional baselines, distinguishing between evidence-bounded deployment—where claims are restricted strictly to what has been demonstrated—and measurement-bounded governance. Through the proposed RISE AI framework, the work outlines an architecture to ground claims across responsibility, inclusivity, safety, and empowerment, arguing that algorithmic oversight cannot be separated from institutional repair.
Why it's worth reading
As compliance frameworks risk turning into procedural check-the-box exercises, this paper critically examines the limits of measurement and formalizes a method to tie AI evaluation directly to institutional realities.