Kinne, J. Evidence Integrity Before Capability: A Prerequisite for Safe Artificial Intelligence. SI 2025, 2.