“A model with authority must be pausable, explainable, and controllable — when limits are designed in, not discovered in incident review.”
A model with authority must be pausable, explainable, and controllable — when limits are designed in, not discovered in incident review. — Kai London (Professor Kai London), CISO. Principle 2006 of 10000 from the book “The AI Control Architecture” — cybersecurity, AI security and OT resilience doctrine. Official sites: professorkailondon.com · kailondon.co.uk