Mar 2026Policy memo
Designing Agent IDs
A policy memo mapping the design space for agent ID systems: functions, existing protocols, private incentives, and ten central design questions
Agent IDs
An open research community building the trust layer for AI agents.
An initiative by Singapore AI Safety Hub (SASH)
Why agent IDs?
Agent IDs identify an agent and its user, carry information about their permissions and enable effective incident response. Services on the internet can require such IDs to assure the trustworthiness, security and accountability of AI agents. No ID, no access.
What we're working on
We want to advance secure and robust agent IDs globally. We define technical protocols, build proof of concepts and conduct policy research.
Our edge? We engage a diverse set of stakeholders to develop truly open agent ID solutions, interoperable with existing private and public identity infrastructure. We work directly with regulators to create the governance processes necessary for broad adoption.
Research
View all- Coming soonPolicy memo
Emergency Shutdown Mechanisms for AI Agents
A policy memo mapping threat models, technical requirements, and governance mechanisms for emergency shutdown of AI agents.
Technical Demo
The demo is an interactive implementation of an agent ID protocol. It details and documents the specific steps each entity takes to allow an AI agent to securely access a web service on behalf of its deployer.
Currently the demo follows a scripted set of steps that show how each component of the agent ID is assembled and used. Next, we will bring each entity to life, letting a viewer inspect network requests and server code as the process runs in real time.
We're building in public — Learn more about the Technical Demo
Join the community
We'd love to shape the future of agent IDs with you—whether you're a researcher, a regulator, an AI developer or an agent deployer.
