Answers are grounded in this paper and cite their sources.
This paper examines a structural gap in Article 12 of the EU AI Act between auditability and verifiability. While the regulation mandates automatic logging for high-risk AI systems, it does not specify mechanisms to independently verify the integrity, completeness, and authenticity of those logs. Drawing on precedents from cryptographic audit logging, Certificate Transparency (RFC 6962), financial market regulation (MiFID II, SEC Rule 17a-4), and digital forensics standards (NIST SP 800-86), the paper identifies five minimum verification properties required for audit logs to function as independently verifiable evidence. The analysis argues that auditability without cryptographic verifiability risks turning regulatory compliance into a trust-based assertion rather than an evidence-based guarantee, and outlines technical and regulatory implications for forthcoming CEN-CENELEC harmonized standards under the EU AI Act.
No takes yet. Share an insight, caveat, or question.
Tokachi Kamimura (2025) studied this question.