GRC Engineer AIUC-1: The First Compliance Framework for AI Agents w/ Rajiv from AIUC and Danny from Schellman
Mar 5, 2026
Rajiv Dattani, co-founder of AIUC and former McKinsey partner focused on AI safety and insurance, and Danny Manimbo, longtime Schellman practitioner and ISO lead in security and compliance, discuss AIUC-1. They cover the six pillars (data, security, safety, reliability, accountability, societal risks). They explain thousands of adversarial simulations, a 100-page audit report plus certification, quarterly updates, and how insurance at Lloyd's aligns incentives.
AI Snips
Chapters
Transcript
Episode notes
Use The 100 Page Report To Speed Vendor Reviews
- Expect a detailed external deliverable: certification plus a ~100-page technical report you can share with buyers.
- Danny said the report replaces roughly 70% of vendor-questionnaire back-and-forth during procurement.
Three Evidence Streams And Thousands Of Tests
- AIUC-1 testing combines three evidence types: adversarial technical testing, technical control review, and policy checks.
- They replay real-world incidents and research attacks, running thousands of adversarial simulations to evaluate behavior.
Insurance Drives Practical Risk Controls
- AIUC-1 is explicitly designed to feed insurers synthetic loss and simulation data so insurers can price AI risk.
- Rajiv traced the idea to historical insurance-driven safety improvements like building codes and airbags.