PILLAR 01 / 04RESEARCH DIRECTION

AI for DPI

Secure interaction between AI agents, identity, credentials and public digital infrastructure.

Read the research questions ↓Collaborate

01 — The basics

How an AI agent acts on someone’s behalf

PersonGrants consent and sets limits on what the agent may do.
AI agentProves its own identity and the authority it has been given.
DPI serviceVerifies the agent and the delegation before acting.

02 — Research questions

What we are asking

Q1How can an AI agent prove who it is and whom it acts for?
Q2How do people grant, limit and revoke what an agent may do on their behalf?
Q3How can DPI services recognise and resist compromised or malicious agents?

03 — Areas of work

Where the work is focused

01
Agent identity
How agents are identified and linked to the person or organisation they represent.
02
Agent consent and authorization
How permissions are granted, scoped, audited and revoked.
03
Agent security
How services detect compromised agents and limit the harm they can cause.

04 — Whitepaper

Trusted AI agents for digital public infrastructure

A framework for how AI agents identify themselves, carry delegated authority and are held accountable when they use identity, credential and payment services.

STATUS: Draft for comment
VERSION: v0.3 · Sample
LENGTH: ~40 pages

01Why agents change DPI
02Agent identity and delegation
03Consent, scope and revocation
04Security and abuse resistance
05Recommendations for DPI operators

Download draft PDF ↓Comment on the draft

05 — Technology roadmap

From research to open technology

CompleteIn progressPlanned

Q4 2026
Threat model and requirements
Map how agents interact with DPI and where they can be abused.COMPLETE
Q1 2027
Agent identity prototype
Agents present verifiable identity and delegation to a test service.IN PROGRESS
Q3 2027
Consent and revocation protocol
Scoped permissions that people can grant and withdraw.PLANNED
Q1 2028
Pilot with a DPI service
Test the framework with a live public service partner.PLANNED
Q3 2028
Open framework v1.0
Specification, reference code and conformance tests.PLANNED

END GOALAn open framework for trusted AI-agent interaction with DPI.SAMPLE DATES

06 — Demos

See the technology working

07 — Playbooks

Practical guides for putting it to use

Step-by-step guidance for governments, service providers and builders adopting this pillar’s work.

FOR DPI OPERATORSSAMPLE
Accepting AI-agent requests safely
How to verify an agent, check its delegation and log what it did.
Download ↓ Accepting AI-agent requests safely24 pages · PDF
FOR PRODUCT TEAMSSAMPLE
Designing consent for agent delegation
Patterns for granting, limiting and revoking agent permissions.
Download ↓ Designing consent for agent delegation18 pages · PDF

08 — Papers

Publications from this pillar (4)

Sort

2026
Conference paperSAMPLE
A. Author, B. Author, C. AuthorVenue to be added · 2026
PDF ↓ Delegated authority for AI agents in public digital services DOI ↗ Delegated authority for AI agents in public digital services
PreprintSAMPLE
A. Author, B. Author, C. AuthorVenue to be added · 2026
PDF ↓ A threat model for AI agents interacting with digital identity systems DOI ↗ A threat model for AI agents interacting with digital identity systems
2025
Journal articleSAMPLE
A. Author, B. Author, C. AuthorVenue to be added · 2025
PDF ↓ Scoped and revocable permissions for agent access to DPI DOI ↗ Scoped and revocable permissions for agent access to DPI
Technical reportSAMPLE
A. Author, B. Author, C. AuthorVenue to be added · 2025
PDF ↓ Agents and DPI: a research agenda for the Asia-Pacific DOI ↗ Agents and DPI: a research agenda for the Asia-Pacific

Entries marked SAMPLE show the layout only. Real publications will be added through the CMS.

Collaborate

Working on AI agents and DPI? Talk to us.

dpi@ait.asia

AI for DPI

How an AI agent acts on someone’s behalf

What we are asking

Where the work is focused

Agent identity

Agent consent and authorization

Agent security

Trusted AI agents for digital public infrastructure

CONTENTS

From research to open technology

Threat model and requirements

Agent identity prototype

Consent and revocation protocol

Pilot with a DPI service

Open framework v1.0

See the technology working

Agent delegation walkthrough

Revoking an agent mid-task

Agent identity inspector

Practical guides for putting it to use

Accepting AI-agent requests safely

Designing consent for agent delegation

Publications from this pillar (4)

Delegated authority for AI agents in public digital services

A threat model for AI agents interacting with digital identity systems

Scoped and revocable permissions for agent access to DPI

Agents and DPI: a research agenda for the Asia-Pacific

Get involved in AI for DPI

Working on AI agents and DPI? Talk to us.

AI for DPI

How an AI agent acts on someone’s behalf

What we are asking

Where the work is focused

Agent identity

Agent consent and authorization

Agent security

Trusted AI agents for digital public infrastructure

CONTENTS

From research to open technology

Threat model and requirements

Agent identity prototype

Consent and revocation protocol

Pilot with a DPI service

Open framework v1.0

See the technology working

Agent delegation walkthrough

Revoking an agent mid-task

Agent identity inspector

Practical guides for putting it to use

Accepting AI-agent requests safely

Designing consent for agent delegation

Publications from this pillar (4)

Delegated authority for AI agents in public digital services

A threat model for AI agents interacting with digital identity systems

Scoped and revocable permissions for agent access to DPI

Agents and DPI: a research agenda for the Asia-Pacific

Get involved in AI for DPI

Working on AI agents and DPI? Talk to us.