A research programme on AI compliance, governance and accountability
We study how AI actions can be recorded, checked, reviewed and challenged—and how those technical capabilities connect to compliance requirements, institutional authority and ethical constraints.
A condition of governability: authorized people can inspect execution, verify evidence, question decisions and trace responsibility. Privacy, proportionate disclosure and access control are part of that condition.
Two connected questions: how can execution records be verified, and what data exposure should platforms record?
How do records become verifiable evidence?
Portable evidence for AI-agent accountability: construct a signed bundle, alter a record, and see what an independent verifier can check.
Interactive notebook · implementations · discussion · ethics
Explore AEGISWhat data was the AI exposed to?
LLM Exposure Monitoring studies how platforms can record data access, API interactions and context exposure using shared event semantics.
Paper · recording schema · reproducible experiments
Explore LEMA broader reading collection on evidence, institutions and human agency. These works retain their own research questions and supporting materials.