AI and Open Banking

Open Banking is particularly well suited to the introduction of AI-initiated payments. It already treats payment initiation as a governed act of delegated authority, built on explicit mandates, scoped authority, and per-action validation, rather than inferring authority solely from credential use.

However, when AI makes decisions autonomously, decision-making moves inside the system, and accountability frameworks designed for human actors and after-the-fact review are no longer sufficient.

This is where Open Banking’s evidentiary model becomes important.

Audit-Ready Artefacts in Open Banking

Open Banking is already strong at evidencing authority and execution. It generates structured, supervisory-grade evidence around payment authority and execution. These artefacts enable reconstruction of who was allowed to act and what occurred:

1. Consent Records

2. Strong Customer Authentication Records

3. Payment Initiation Records

4. Transaction and Event Logs

5. TPP Identity and Licensing Evidence

6. Bank Execution Records

7. Scheme and Dispute Records

Together, these demonstrate that:

• a user authorised a payment

• a regulated party initiated it properly

• a payment was executed as instructed

The Governance Gaps Introduced by AI

When payments are initiated by AI rather than humans, new accountability gaps emerge. Traditional controls show that a payment was authorised and executed, but not whether the AI was authorised to decide, acted within mandate, or triggered the payment under legitimate conditions.

Four areas become especially important:

A) Authority Systems must demonstrate who delegated which decision rights to a specific AI agent, under what mandate and limits.

B) Intent Intent must be demonstrable. Without this, an AI-initiated payment can appear indistinguishable from misuse or error.

C) Constraint It must be demonstrable that the AI agent acted within policy, mandate, and scope at the time each action occurs.

D) Causality A link is required between delegated authority, the AI decision, and the executed payment to evidence a legitimate and auditable chain of events.

These elements extend accountability beyond payment execution to the legitimacy of the decision itself.

A New Actor Within an Existing Framework

The aim is not to change Open Banking, but to enable AI to operate safely and effectively within it.

Existing roles, rules, and regulatory expectations remain intact. What evolves is the ability to apply them to autonomous decision-makers in a machine-verifiable way.

MandateProof provides a governance and evidentiary layer that makes authority, decision context (intent), constraints, and decision-to-execution links machine-verifiable.

This allows AI agents to operate within Open Banking frameworks while preserving existing supervisory expectations.