Skip to content
Atesto
Create account
All articles

AI at work

Singapore tests DNS to track autonomous AI agents

Singapore is testing the internet domain name system to verify AI agents, setting a baseline for how organizations track shadow AI.

Singapore is testing the internet domain name system to verify AI agents.

On November 13, 2026, Biometric Update (opens in a new tab) reported a new initiative from the government of Singapore to adapt the global internet domain name system to identify, track, and verify autonomous artificial intelligence agents. The Infocomm Media Development Authority partnered with infrastructure companies including GoDaddy, Cloudflare, Infoblox, and Google for this project. For technology and compliance officers, the initiative highlights a shift in how organizations track machine identities. Relying on manual software inventories falls short when employees deploy independent agents that operate continuously in the background.

Who it applies to

The immediate exploration involves the Infocomm Media Development Authority and its industry partners testing technical architectures. Concurrently, the Government Technology Agency of Singapore is developing a shared infrastructure called GovHarness for building and operating artificial intelligence agents across the public service. While these are government and industry initiatives, they establish a baseline for enterprises operating in Singapore or interacting with its digital infrastructure.

The concepts apply directly to enterprise technology teams managing network access and compliance officers responsible for enforcing acceptable use policies. When employees connect unapproved artificial intelligence tools to their corporate accounts, they often authorize autonomous agents to read emails, alter records, or trigger transactions. This practice creates a new layer of shadow AI. Even if an employee deploys an unapproved tool, the organization remains accountable for the actions the agent takes on its network. The infrastructure tracking initiative shows that authorities expect organizations to know exactly which machine identities operate within their perimeters.

What it requires

The working group presented a demonstration of a Chain of Trust architecture in October 2026. This architecture verifies agent identities during transactions across independent platforms. The goal is to ensure that organizations can prove which human or corporate entity controls a specific automated process. Senior Minister of State for Digital Development and Information Jasmin Lau stated that the objective is ensuring that "behind every agent, there is always someone who does."

This exploration reflects a broader expectation regarding enterprise visibility. Traditional shadow technology involved employees using unapproved software applications. Information technology departments could often detect these through network traffic analysis or expense reports. Autonomous shadow AI is more complex. Employees now delegate workflows to autonomous tools that interact directly with enterprise software interfaces. These agents operate continuously, making decisions and executing commands without secondary human approval.

Singapore is also considering formal registration mechanisms for these tools. Member of Parliament Tin Pei Ling proposed a national registry modeled on Singpass, the digital identity system of the country. The proposal includes a concept of biometric passporting for artificial intelligence agents, built on open, interoperable standards. If national governments build formal registration and passporting for automated systems, enterprise frameworks face expectations for similar internal controls. A static spreadsheet listing approved vendors cannot capture the dynamic nature of autonomous sub-agents.

Furthermore, the GovHarness infrastructure being developed for the public sector emphasizes concrete technical requirements. Senior Minister of State Tan Kiat How stated that the system supports deployment, connections to tools, and capabilities for monitoring and evaluation. It includes common guardrails designed to help developers observe agent actions and identify unexpected behavior. Tan noted that building quickly is not the same as deploying safely. The approach is to build standards directly into the platforms people use, rather than expecting every officer to become an expert in hosting and cybersecurity. The goal is to translate the principles of the Model AI Governance Framework for Agentic AI into technical requirements. Organizations face the expectation to adopt similar architectures to monitor their own environments. The standard shifts from merely listing the tools employees use to actively verifying the identity, origin, and authorization level of every automated request hitting their servers.

The dates

The Infocomm Media Development Authority and its industry partners presented their initial demonstration of the Chain of Trust architecture in October 2026. The details of the domain name system exploration, the GovHarness infrastructure, and the proposals for a national registry were discussed in Parliament and reported on November 13, 2026. Because this is an exploratory initiative and a set of proposals rather than a finalized regulation, there are no binding compliance deadlines currently in force.

What to check now

  • Review network logs for unusual application programming interface activity that indicates automated agent operations rather than standard human interaction.
  • Update acceptable use policies to explicitly define the rules for deploying autonomous artificial intelligence agents and automated task execution.
  • Map which internal systems allow external applications to read, write, or delete data without requiring a secondary human authorization prompt.
  • Evaluate current identity and access management configurations to ensure that machine identities and agent accounts are strictly separated from human user credentials.
  • Assess the current software inventory process to determine if it can track dynamic, autonomous sub-agents operating within approved platforms.

The shift from human identity verification to autonomous agent verification requires a structural update to enterprise security models. As governments and infrastructure providers build frameworks to track automated systems at the routing layer, organizations face the expectation to identify exactly what operates on their behalf. Establishing clear boundaries for employee technology use is the first step in maintaining control over automated enterprise environments.

Sources

  1. Biometric Update (opens in a new tab) · www.biometricupdate.com

This article is general information, not legal advice. For your company's situation, talk to a qualified lawyer.

This article was written with AI and checked, sentence by sentence, against the sources linked above.