Authority & governance
SAF-G
A framework governing when an action proposed by AI may proceed, with explicit permission checks and recorded outcomes.
For enterprise AI and platform teams
THE PURPOSE
A specific problem.
A defined evaluation.
An AI proposal needs a defined route to an authorised action. SAF-G focuses on the permissions, boundaries and evidence that make that decision inspectable.
A USE CASE TO EXAMINE
Evaluate a workflow where an agent proposes a change to a business system. Examine the approved scope, a changed request and the record available to the person responsible for the outcome.
What to review together
- Proposal identity and permission scope
- Boundary changes and refusal behaviour
- Decision records and independent verification
The demonstration, build and supporting evidence are agreed for the specific evaluation. Readiness is assessed for that scope.
THE ENGINEERING PRACTICE
Its own estate.
The same discipline.
Each estate has its own codebase, workstreams, technical documentation, validation records, operating and recovery procedures, and controlled release process.
An evaluation can examine the relevant architecture, validation evidence, recovery behaviour and integration boundary. Detailed technical material is shared within an agreed review scope.
EVALUATION · INTEGRATION · LICENSING
Start with your use case.
Tell me what you want to achieve, the environment involved and what you need to establish before making a decision.
Discuss SAF-G ↗