From Protocols to Evidence: Bounded Claims for AI in Service of the Common Good

Analysis

Why This Paper Matters

AI governance has largely focused on translating principles into protocols—roles, requirements, records, oversight, and assessment. The EU AI Act, NIST AI RMF, ISO/IEC 42001, and assurance practices exemplify this move. Yet, as Chawla and Benanti argue, these protocols leave harder questions unanswered: what they actually establish, whose power they leave untouched, and where measurement must stop. This paper matters because it shifts attention from compliance mechanics to the institutional ruptures that AI systems enter and potentially repair, compound, substitute for, or conceal.

By introducing a rupture test, the authors force evaluators to assess not only the AI system but also the institutional baseline it is deployed into. This dual focus is a significant departure from system-centric evaluation. It acknowledges that AI can reveal where institutions have already failed to provide responsiveness, belonging, care, and accountability. Responsible AI, then, is not merely a technical or regulatory problem but also a matter of institutional repair and moral judgment.

The paper also draws on Pope Leo XIV's Magnifica Humanitas to ground its argument in a broader moral frame centered on dignity, technological power, and the common good. This integration of moral philosophy with AI governance protocols is unusual and provides a richer normative foundation than typical technical or legal analyses.

Technical Contributions

• Rupture test: A method that links institutional baselines to system evaluation, requiring assessment of both the AI system and the institutional failure it encounters.
• Evidence-bounded deployment: A principle that limits claims about an AI system to what has actually been evaluated, preventing overreach beyond empirical evidence.
• Measurement-bounded governance: A complementary principle that records constraints which favorable evidence cannot override, ensuring that certain moral or political limits remain non-negotiable.
• RISE AI architecture: A framework for making bounded, evidence-based claims about Responsibility, Inclusivity, Safety, and Empowerment.
• Integration of protocols and moral frame: Connects existing governance instruments (EU AI Act, NIST AI RMF, ISO/IEC 42001) with a moral framework centered on the common good.

Results

The paper reports no empirical metrics, benchmarks, or quantitative comparisons. Its results are conceptual: the rupture test, the distinction between evidence-bounded deployment and measurement-bounded governance, and the RISE AI architecture. These are proposed as analytical tools rather than validated instruments. The authors do not provide case studies or implementation data to demonstrate effectiveness.

Significance

The broader impact of this work lies in its potential to reshape how AI practitioners and policymakers think about evaluation and governance. By insisting that claims be bounded by evidence and that governance record constraints that evidence cannot override, the paper offers a vocabulary for resisting the tendency to overclaim AI capabilities and to ignore institutional context. The rupture test could influence assurance practices by requiring institutional baselines as part of system evaluation. The RISE AI architecture provides a structured way to make bounded claims across four dimensions. Ultimately, the paper argues that responsible AI requires better engineering, institutional repair, and continued moral and political judgment—a tripartite demand that challenges purely technical or purely regulatory approaches.