What You Need to Know (And Why You Should Care) About AI Governance
summary
Argues that the governance unit shifts from a model to the whole workflow and that engineering controls (sandboxes, permissions, logging, eval suites) are the real mitigation. Recommends NIST RMF as the practical lens (NIST RMF: govern, map, measure, manage), notes Microsoft maps to it, and flags the EU AI Act change (EU AI Act: transparency obligation from Sept; high‑risk obligations delayed). Gives a tiny worked example — an AI support bot with an “AI” label, bias checks, and an eval CI test — and warns about ethical debt, gold‑drift, and agent modularity.
AI-written summary · about
