Herramienta gratuita

Checklist de preparación SAFR — Evaluación de brechas para la gobernanza de la IA agéntica

35 preguntas basadas en el white paper de MAS BuildFin.ai. Se puntúan en tu navegador. Sin registro.

Safeguards for Agentic Finance at Runtime (SAFR) es el white paper que la iniciativa BuildFin.ai de la MAS publicó en julio de 2026, elaborado junto con Ant International, Circle, HSBC, J.P. Morgan Chase, Manulife, Mastercard, OCBC y Visa. Describe una capa de gobernanza en tiempo de ejecución situada entre un agente de IA y los sistemas sobre los que actúa, y fija un único estándar para cada acción propuesta: «ninguna acción agéntica llega a ejecutarse sin haber sido declarada, autorizada y evaluada».

Este checklist traduce ese estándar en 35 preguntas repartidas en ocho secciones, siguiendo los cuatro componentes de tiempo de ejecución del white paper —Identidad del Agente, Repositorio de Controles, Motor de Disposición, Registro de Auditoría— y su tratamiento de mandatos, sobres de gobernanza, escalado y despliegue. Cada pregunta nombra la evidencia que un evaluador solicitaría, de modo que responder Sí significa que puedes entregar el artefacto esta semana.

Responde Sí, Parcial o No. Las puntuaciones se calculan en tu navegador, por sección y en conjunto. El conjunto completo de preguntas está en esta página; el registro por correo solo se aplica a la descarga de la hoja de cálculo editable. Si el marco es nuevo para ti, empieza con SAFR, explicado. Una vez que tengas tu puntuación, la guía de implementación de SAFR explica cómo cerrar cada brecha, y la página implementación de SAFR muestra cómo KLA Control Plane se alinea con el marco.

SAFR readiness: 0%

Assessed 0/35 · 0 gaps

Agent inventory and identity

SAFR's Agent Identity component binds each proposed action to a recognized, registered agent, verified against a registry entry before any other evaluation proceeds.

Do you maintain a complete inventory of every AI agent able to initiate actions against production systems — payments, trading orders, credit approvals, regulatory filings, claims?

Evidence: An export of the inventory listing each agent, the systems it acts on, and its current release state.

SAFR A1

Does every agent in that inventory have a named human owner accountable for its behavior?

Evidence: The registry records showing an owner field populated with a named individual for each agent.

SAFR A2

Is every agent versioned, so any past action can be traced to the specific agent version that produced it?

Evidence: The version history for one agent, and a sample action record referencing the version that acted.

SAFR A3

Where an agent could appear in more than one registry — an internal registry, a payment network's agent registry, an inter-institutional directory — have you determined which registry is authoritative?

Evidence: A written designation of the authoritative registry per agent class, or confirmation that all agents operate in a closed-loop environment against a single internal registry.

SAFR A4

Is agent identity verified against the registry at runtime, before each proposed action is evaluated, with verification failures rejected and logged?

Evidence: A log entry for a rejected action showing failed identity verification as the recorded basis.

SAFR A5

Mandates and authorization

The paper defines a mandate as "the mechanism through which a user defines the bounds of the authority delegated to an agent" — explicit, structured, and machine-readable.

Is the authority delegated to each agent recorded as an explicit, machine-readable mandate stating what the agent may do, within what limits, and under what conditions?

Evidence: One production mandate in the machine-readable form the runtime actually consumes, with the issuing principal identified.

SAFR B1

Are the action types each agent may initiate enumerated in its mandate, with a scope the agent cannot extend through its own reasoning or inference?

Evidence: The mandate's permitted action-type list, plus a test or log showing an out-of-scope action rejected at the checkpoint.

SAFR B2

Is delegation depth defined — who may delegate authority to whom, and how many links a delegation chain may carry?

Evidence: The authorization rules covering delegation, with the maximum chain depth stated.

SAFR B3

Does every mandate carry a validity period, with actions proposed after expiry rejected until the mandate is renewed?

Evidence: A mandate showing its validity window, and the outcome recorded for an action proposed after expiry.

SAFR B4

Can a mandate be revoked or suspended with immediate effect, applying from the next proposed action?

Evidence: The revocation procedure, and a test record showing an action under a revoked mandate rejected.

SAFR B5

Controls repository

The Controls Repository is the institution's configurable rulebook, drawn from organizational policies, regulatory requirements, product rules, and user-provided mandates.

Are the controls governing agent actions held in one designated repository, and does the runtime evaluation read from it?

Evidence: The repository location, and a configuration or trace showing the evaluation path retrieves controls from it.

SAFR C1

Is every control versioned, so any past decision can be traced to the exact control version in force when it was made?

Evidence: The version history for one control, and a decision record citing the version applied.

SAFR C2

Is publishing a control change a governed release, reviewed and approved by a second person before it takes effect?

Evidence: The approval record for the most recent control change, showing author and approver as different people.

SAFR C3

Do the controls cover all four sources SAFR names — organizational policies, regulatory requirements, product rules, and user-provided mandates?

Evidence: One example control per source, mapped to its origin in the repository.

SAFR C4

Disposition engine

The Disposition Engine evaluates each in-scope action against the retrieved controls, producing "a defined, binding outcome for every proposed action calibrated to the specific risk it presents."

Is every in-scope proposed action evaluated deterministically against the controls before it executes — identical envelope, mandate, and control version producing an identical outcome?

Evidence: An execution trace showing the checkpoint positioned ahead of the action with no bypass route, plus a repeat-evaluation test confirming determinism.

SAFR D1

Can the evaluation return all four SAFR dispositions — Deny, Escalate, Auto-Execute, and Observe?

Evidence: One logged decision of each type, or configuration showing all four outcomes reachable.

SAFR D2

Does every outcome record the specific rules applied and a machine-readable reason for the result?

Evidence: A decision record showing rule identifiers and reason codes alongside the outcome.

SAFR D3

Are the thresholds that assign outcomes calibrated on the factors SAFR names — action reversibility, financial materiality, customer impact severity, regulatory sensitivity, and novelty?

Evidence: The calibration parameters for one agent workflow, with the rationale recorded for each threshold.

SAFR D4

If the evaluation service is unreachable or errors, is the proposed action blocked by default?

Evidence: A failover or chaos test record showing an action held or rejected while evaluation was unavailable.

SAFR D5

Governance envelope and lineage

Before an action executes it is packaged in a Governance Envelope — the action, the trace of how the agent arrived at it, and the context needed to assess it — "treated as a document to be authenticated against its origin."

Does each proposed action carry a trace of the actual steps the agent executed in arriving at it — tool calls made, data retrieved, checks performed?

Evidence: A captured action trace for one production or pilot action, showing the ordered tool-call sequence.

SAFR E1

Does the envelope carry context metadata: the agent identity, the applicable mandate, current account or system state, and operative policy constraints?

Evidence: A sample envelope with all four metadata fields populated.

SAFR E2

Is the envelope authenticated against its origin, so a trace and action declared together by a compromised agent would be caught before evaluation?

Evidence: A description of the authentication mechanism — for example, envelope assembly at an interception point in the execution path — and the adversarial test that exercised it.

SAFR E3

Can the full lineage of any completed action be reconstructed afterward — every step, decision, and data access, in order, from the stored record alone?

Evidence: A lineage reconstruction for one past action, produced without reference to application logs or to the agent itself.

SAFR E4

Escalation operations

SAFR names three operational dimensions for escalation — volume, turnaround, and reviewer authority — and warns that ad hoc escalation yields "the appearance of human oversight without the substance of it."

Does an escalated action pause and route to a named, authorized reviewer as a structured decision request with the full action context attached?

Evidence: One escalation record showing the held action, the assigned reviewer, and the context packaged with the request.

SAFR F1

Do reviewers hold clear authority to approve, modify, or decline, with their decisions carrying the same institutional weight as the original agent decision?

Evidence: A role definition granting reviewers that authority, with their decisions binding on the workflow.

SAFR F2

Is maker-checker separation enforced by the system, barring the initiator of an action from approving it?

Evidence: A test record showing a self-approval attempt rejected by the platform.

SAFR F3

Does every escalation carry a timeout window with a defined default when no decision arrives in time — block, or escalation to a senior reviewer?

Evidence: The timeout and default outcome configured per decision class, with coverage rules for overnight and weekend hours.

SAFR F4

Have you sized expected escalation volume against actual reviewer capacity, including peak periods?

Evidence: A volume forecast at expected traffic set beside the reviewer roster, with turnaround data from a pilot or production period.

SAFR F5

Audit and evidence

The Audit Log is the record every other component feeds — "the authoritative record, independent of any party with an interest in how events are characterised after the fact."

Is every governance decision — all four outcomes, including Auto-Execute and Observe — written to an append-only, tamper-evident log?

Evidence: The log design (hash chaining or equivalent) and one sample entry per outcome type.

SAFR G1

Does each entry capture what SAFR requires: the envelope as submitted, the mandate checked against, the outcome, the specific rules applied, the basis for the outcome, and the time elapsed at each stage?

Evidence: One complete log entry showing all six elements.

SAFR G2

Can you export the governance record for a given agent, period, or case in a form an auditor or supervisor can work with?

Evidence: One export produced end to end, with the elapsed time to produce it.

SAFR G3

Can a third party verify the integrity of an exported record without access to your systems?

Evidence: An offline verification run against an exported bundle, with the verifier's output attached.

SAFR G4

Deployment pattern

SAFR defines two integration patterns — native instrumentation and gateway interception — and requires per-action authorization across multi-step work.

Have you chosen an integration pattern for each part of the agent estate: native instrumentation for agents you build, gateway interception for existing and third-party agents?

Evidence: The pattern decision recorded per agent class, with the rationale for each.

SAFR H1

Does a coverage plan establish governance over the existing agent estate first, with native instrumentation following for new builds?

Evidence: The rollout plan listing every currently ungoverned agent and its target coverage date.

SAFR H2

In multi-step workflows, is each action authorized independently, with an Auto-Execute or Observe outcome at one step carrying no authority into the next?

Evidence: A multi-step run record showing a separate evaluation and outcome at every step.

SAFR H3

Prioritised gaps

35 item(s) to address

Do you maintain a complete inventory of every AI agent able to initiate actions against production systems — payments, trading orders, credit approvals, regulatory filings, claims? — Not assessed
Does every agent in that inventory have a named human owner accountable for its behavior? — Not assessed
Is every agent versioned, so any past action can be traced to the specific agent version that produced it? — Not assessed
Where an agent could appear in more than one registry — an internal registry, a payment network's agent registry, an inter-institutional directory — have you determined which registry is authoritative? — Not assessed
Is agent identity verified against the registry at runtime, before each proposed action is evaluated, with verification failures rejected and logged? — Not assessed
Is the authority delegated to each agent recorded as an explicit, machine-readable mandate stating what the agent may do, within what limits, and under what conditions? — Not assessed
Are the action types each agent may initiate enumerated in its mandate, with a scope the agent cannot extend through its own reasoning or inference? — Not assessed
Is delegation depth defined — who may delegate authority to whom, and how many links a delegation chain may carry? — Not assessed
Does every mandate carry a validity period, with actions proposed after expiry rejected until the mandate is renewed? — Not assessed
Can a mandate be revoked or suspended with immediate effect, applying from the next proposed action? — Not assessed
Are the controls governing agent actions held in one designated repository, and does the runtime evaluation read from it? — Not assessed
Is every control versioned, so any past decision can be traced to the exact control version in force when it was made? — Not assessed
+ 23 more in the export.

Export your scored assessment and prioritised gap list. Everything stays in your browser — nothing is uploaded.

Cómo interpretar tu puntuación

Sí = 2, Parcial = 1, No = 0. Las puntuaciones son porcentajes de los puntos disponibles, por sección y en conjunto.

Brechas fundamentales

0–39 %

Las puntuaciones en este rango señalan estructuras ausentes: un inventario de agentes con propietarios nombrados, mandatos legibles por máquina y un punto de control que evalúa las acciones antes de que se ejecuten. Mientras no existan, la calibración de disposiciones y las herramientas de evidencia no tienen a qué aferrarse. Empieza por las secciones A y B: inventaría cada agente capaz de iniciar una acción, nombra un propietario para cada uno, y documenta su autoridad delegada como un mandato. La guía de implementación de SAFR ordena el trabajo.

Cobertura parcial

40–75 %

Las estructuras principales existen, con brechas que un evaluador encontraría en la primera hora: controles dispersos en el código de los agentes, escalados sin plazos por defecto, un registro de auditoría que ninguna parte externa puede verificar. Cierra primero la sección peor puntuada; cada sección corresponde a un frente de trabajo en la guía de implementación. Vuelve a puntuar después de cada frente de trabajo.

Lista para piloto

76–100 %

Las estructuras que describe SAFR están sustancialmente implementadas. El trabajo restante es de calibración: umbrales de disposición ajustados con tráfico real, volumen de escalado dimensionado a la capacidad de los revisores, exportaciones de evidencia probadas con los auditores que las van a consumir. Un piloto gobernado de cuatro semanas sobre un flujo de trabajo de un agente es un siguiente paso realista; la página implementación de SAFR describe uno, semana a semana.

Lleva la puntuación a una sesión de trabajo

Una revisión de brechas de 30 minutos contrasta tus secciones peor puntuadas con KLA Control Plane, que implementa el patrón SAFR y ya está desplegado. La revisión toma tus puntuaciones por sección y asocia cada brecha a un control concreto.

Reserva una revisión de brechas SAFR de 30 minutos

Descarga el checklist editable

Las mismas 35 preguntas en una hoja de cálculo: una fila por pregunta con su nota de evidencia, además de columnas de Estado, Responsable y Notas, y puntuación por sección. Pensada para revisiones internas de brechas.

Se usa para enviarte el archivo y, como máximo, un único seguimiento.

Sobre el marco

SAFR —Safeguards for Agentic Finance at Runtime— se publicó en julio de 2026 por la iniciativa BuildFin.ai de la MAS, elaborado junto con ocho miembros del sector, entre ellos HSBC, J.P. Morgan Chase, Mastercard y Visa. Es un enfoque de referencia para que las instituciones lo implementen dentro de su propia infraestructura. El white paper indica que «no constituye orientación regulatoria ni expectativas de supervisión».

SAFR, explicado — el marco completo

Qué produce una buena puntuación

Una implementación bien puntuada aquí puede entregar a un evaluador cuatro artefactos bajo demanda: un registro de decisión con motivos legibles por máquina, una traza reconstruible de los pasos que realmente siguió el agente, un registro de aprobación con el nombre del revisor y su justificación, y una exportación de evidencia que un tercero puede verificar sin conexión. Las secciones D a G evalúan cada uno de estos elementos.

Implementación de SAFR, componente a componente

Cerrar las brechas

KLA Control Plane implementa el patrón SAFR y ya está desplegado: Agent Registry para la identidad, Policy Builder para el repositorio de controles, KLA Policy Engine para las disposiciones, Audit Trail para el registro de evidencia. Una sesión de treinta minutos recorre una ejecución gobernada de principio a fin: política, decisión, escalado, evidencia sellada.

Implementación de SAFR con KLA Control Plane Reserva una revisión de brechas SAFR de 30 minutos

Preguntas

Preguntas frecuentes

¿Es esta una evaluación oficial de la MAS?

No. Se trata de una autoevaluación construida a partir de las estructuras que define el white paper de SAFR, sin ningún estatus oficial. El propio SAFR es una referencia del sector —el white paper indica que «no constituye orientación regulatoria ni expectativas de supervisión»— y cada institución sigue siendo responsable de determinar cómo se alinea su implementación con las expectativas de supervisión aplicables y sus requisitos internos de gobernanza. KLA es independiente de la MAS y de BuildFin.ai.

¿Cuándo debemos responder Parcial?

Responde Parcial cuando la capacidad cubra solo parte del parque de agentes, funcione únicamente en un piloto, o exista sin la evidencia nombrada bajo la pregunta. La nota de evidencia es la prueba: Sí significa que podrías entregar el artefacto a un evaluador esta semana.

¿Quién debería completar el checklist?

Tres roles lo cubren: quien es propietario de la plataforma de agentes, quien es propietario de los controles de cumplimiento relevantes, y quien dirige la operación de revisión. Cada pregunta se responde con Sí, Parcial o No, así que una primera pasada es rápida; las notas de evidencia son donde realmente se invierte el tiempo de revisión. Las puntuaciones se calculan en tu navegador; el registro por correo solo se aplica a la descarga de la hoja de cálculo.

¿Qué debemos hacer con las brechas que revela la puntuación?

Aborda primero la sección peor puntuada. Cada sección corresponde a un frente de trabajo en la guía de implementación de SAFR; vuelve a puntuar después de cada frente de trabajo y el desglose por sección mostrará dónde se movió la puntuación. Una brecha se considera cerrada cuando puedes producir el artefacto de evidencia nombrado bajo su pregunta.

Aviso legal: Autoevaluación frente a Safeguards for Agentic Finance at Runtime, white paper v1.0 (julio de 2026), MAS BuildFin.ai. Los pasajes citados están protegidos por derechos de autor de la Monetary Authority of Singapore. SAFR es una referencia del sector y no constituye orientación regulatoria ni expectativas de supervisión. KLA es independiente de la MAS y de BuildFin.ai, y no está afiliado a, respaldado por, ni certificado por ninguna de las dos.