Who Owns Validity? AI Assessment, Institutional Boundaries, and the Future of Quality Assurance

Generative artificial intelligence (GenAI) challenges the relationship between assessed student work and the capabilities that universities claim to certify. This conceptual paper examines assessment validity as a shared governance problem, rather than primarily a matter of misconduct detection or technology procurement. It synthesises recent evidence from the UK Quality Assurance Agency, research on automated essay marking, and scholarship on validity-centred and process-informed assessment. The paper proposes a Cross-Institutional AI Assessment Compact comprising co-designed validity indicators, reciprocal audit mechanisms and student-centred accountability. It argues that institutional partnerships can provide a practical setting in which to test assessment redesign, align expectations and protect disciplinary autonomy while developing comparable standards of assurance. The proposed compact remains a framework for pilot evaluation, not an empirically validated intervention.