AI Governance
The policies, controls, responsibilities and evidence used to ensure AI systems are developed and operated appropriately.
Quick help: AI governance is the set of policies, controls, responsibilities and evidence used to ensure AI systems are developed and operated appropriately. Governance should not be treated simply as a compliance document written after development -- it should influence architecture and operation.
Governance can include
Model inventory, data/source controls, access controls, evaluation, human approval, guardrails, versioning, audit trails, risk classification, change management, and incident handling.
Evidence
A useful governance system should be able to answer: what system ran, which model, which version, against what data, using which sources, which tools were available, which guardrails applied, how was it evaluated, and who approved it.
KB Sandbox Principle
Governance should be observable in how the system operates, not merely described in policy.