We don't replace MindForge. We make it enforceable, action by action.
MAS Project MindForge gives Singapore financial institutions a handbook for operationalising AI risk management. The KYE MAS MindForge Banking Authority Pack™ turns the handbook's action-boundary guidance into runtime controls: every consequential AI action proceeds only inside the human-oversight mode you declared for it, only within the authority you recorded for that AI, under a named accountable officer — and emits a signed record you can replay for MAS.
A handbook tells you what good looks like. It does not stop a bad action at the moment it happens.
- Oversight is a posture on paper. The handbook asks you to choose human-in-the-loop, human-over-the-loop, or human-out-of-the-loop per AI use — but a policy document does not gate the action when it runs.
- The AI inventory says what exists, not what is allowed. Knowing an AI is in use is not the same as enforcing what it is authorised to do.
- Third-party and vendor AI widen the surface. An external model acting inside your boundary needs the same authority discipline as your own.
- Monitoring and change management need evidence. When MAS or internal audit asks why an action proceeded, reconstructed notes are contestable.
- Accountability must attach to a person. A consequential AI action that no named officer stands behind is a gap, not a control.
A signed Evidence Pack™ on every consequential action — verifiable from public keys alone.
- Oversight mode enforced, not assumed. A human-in-the-loop action cannot reach finality without a signed approval; a human-over-the-loop action logs every override and review; a human-out-of-the-loop action proceeds only inside its pre-approved scope.
- An AI action-authority inventory. Every AI is recorded with what it is authorised to do; an action outside that authority is refused at the boundary.
- Escalation before finality. An out-of-mode or out-of-scope action returns REQUIRE_APPROVAL and routes to the named accountable officer before it can become final.
- External AI under the same discipline. Vendor and third-party AI act only inside the authority recorded for them.
- Replay-provable for MAS. Every consequential action emits a signed provenance pin binding the model, the inputs, the oversight mode, and the named authority — replayable offline.
Governance at the moment of the action — across every consequential AI decision.
Every consequential AI action flows through the same gate. The pipeline binds to canonical KYE Protocol™ envelopes — no bespoke shapes per institution.
- 1 — Action proposed. An AI agent proposes a consequential action — an approval, a payment, a trade, a credit decision, a customer communication.
- 2 — Oversight-mode gate. The gate checks the action against the declared human-oversight mode for that use. No resolved mode, no action.
- 3 — Authority check. The action is checked against the AI's recorded action-authority — including for external and vendor AI. Outside authority, it is refused.
- 4 — Escalation where needed. An out-of-mode or out-of-scope action returns REQUIRE_APPROVAL and routes dual-channel to the named accountable officer before finality.
- 5 — Named accountability. The action is bound to a named accountable officer; accountability never transfers to the AI.
- 6 — Evidence Pack™ sealed. The action emits a signed provenance pin binding the model, the inputs, the oversight mode, and the named-authority outcome, WORM-retained.
- 7 — Replay for MAS / audit. The Evidence Pack™ is verifiable offline from public keys alone, and a change to an AI's authority is itself an evidenced, replayable event.
Bound to the MAS MindForge action-boundary subset.
The Pack binds the canonical KYE™ artefact set to the action-boundary considerations of the MindForge handbook. Every claim resolves to a control row on the bound framework. We operationalise the handbook — we do not replace it, and we do not govern your governance operating model or judge whether the AI's output is correct.
| MindForge block | Control area | Pack coverage |
|---|---|---|
| Scope & AI oversight | Human-oversight modes enforced + AI action-authority inventory | full |
| AI risk management | Escalation before finality + third-party / vendor AI authority | full / partial |
| AI lifecycle management | Deployment controls at the gateway + monitoring/change as Evidence Packs | full |
| Enablers | Named accountability at the action boundary | full |
Full coverage means every action-boundary requirement in the control area is bound to a canonical KYE™ rule, dictionary term, or Evidence Pack™ contract. Partial coverage means the bound surface satisfies the control area when paired with the institution's own attestation. The handbook considerations outside the action boundary — your governance operating model, model-development methodology, and the correctness of the AI's output — stay yours (honest scope).
Run a pilot on synthetic data — then deploy with confidence.
The MAS MindForge Banking Authority Pack™ ships in three tiers: a fixed-scope oversight-mode readiness diagnostic, a full pilot on synthetic data, and an annual enterprise entitlement. Pricing is value-based and disclosed under NDA to qualified applicants.