AI Runtime Security, Agentic Zero Trust & Oversight

New York’s RAISE Act is in effect, requiring registration, safety-framework and transparency disclosures, and 72-hour critical-incident reporting for large frontier developers; enforcement remains untested. NIST architectures and practical AI RMF guidance emphasize zero-trust controls, accountable ownership, inventories including shadow AI, continuous risk management, action lineage, reversibility, and safe containment. MIT researchers cataloged 2,206 safeguards across 47 frameworks and found gaps in direct system safeguards and autonomous-agent oversight; the resource supports discovery, not proof of efficacy.