Autonomous Audit of AI Governance Architecture

Puntos clave

• The aim is to establish a governance property that ensures sound, verifiable conclusions from distributed AI records, even without complete evidence.
• Proposes a novel concept called Examination Validity for auditing AI governance records.
• Extends traditional governance frameworks to include cross-organizational auditability.
• Illustrates the use of cryptographic techniques to enable verifiable evidence and governance attestations.
• Demonstrates enhanced trustworthiness in AI governance investigations across organizational boundaries.
• Shows that cryptographically verifiable evidence can effectively support audits without extra governance layers.
• Indicates that even partial evidence can lead to reliable conclusions through the defined governance framework.