The Hartstone AI Governance Framework

Runtime Authority for Machine-Initiated Action

Most organizations govern AI on two planes. Information governance decides what a system may read, and model governance assesses how it reasons. Neither operates at the moment an AI system acts in the real world.

When an agent sends the transfer, deletes the record, or signs the message, both planes have already finished their work. This framework closes that gap with a third plane, runtime authority, a deterministic control point between intent and execution.

• ◆The three-plane governance framework: information, model, and runtime authority
• ◆The four-phase governance loop: Diagnose, Decide, Transform, Verify
• ◆An implementation guide: action tiers, authority matrix, and audit event specification
• ◆Three fillable adoption templates, ready to complete during rollout
• ◆A 30/60/90-day rollout plan, from inventory to a closed loop