evidoria

← Back to browse

Good practice Imported

GovTech's AI Agent Registry — Singapore Extends Oversight to 150,000 Public Officers' AI Agents

Singapore · Singapore · See the Singapore profile · See the Singapore profile

Evidence: Descriptive / self-reported Top 46% 60/100 · Ask Evidence Copilot about this practice

GovTech Singapore announced in June 2026 a registry tracking ownership and permitted actions of AI agents used by its 150,000 public officers, building on Pair, its assistant already used by more than half the workforce — piloting now, wider rollout later in 2026.

150,000 officers
Public officers within scope of the AI Assistant Desk/registry (2026)
GovTech's AI Agent Registry — Singapore Extends Oversight to 150,000 Public Officers' AI Agents

Details

Maturity
Pilot
Promoter
GovTech Singapore (Government Technology Agency)
Period
2026 (announced June 2026, phased rollout)
Keywords
public administration, digital government, AI governance, internal operations

Context

GovTech Singapore, the government's digital-transformation agency, runs Pair, a generative-AI assistant already used regularly by more than half of the country's 150,000 public officers.

Objectives

Build a registry of AI agents, as part of an "AI Assistant Desk" suite, that tracks who owns each AI agent and what actions it can take — addressing the risk that agentic AI can act and decide at machine speed without a human checking each step.

Activities

Announced 2 June 2026. The design combines customisable rules, a set of sanctioned AI tools and a registry, plus hygiene checkers that block actions such as unauthorised file deletion or external emails. The registry was being piloted with a subset of officers at announcement, with wider rollout planned for later in 2026.

Results

As of the available reporting, no count of registered agents or measured oversight outcomes has been published — it is a governance design in early rollout, building on an already-adopted assistant, rather than an evaluated, mature control.

Conclusions

Its effectiveness at preventing AI-agent misuse across a 150,000-strong workforce remains to be demonstrated.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Short (< 1 year)
Staffing & skills
GovTech engineering and AI-governance teams, named executive sponsor (GovTech CEO Goh Wei Boon)

Conditions for success

  • a registry that reliably tracks agent ownership and permitted actions
  • hygiene checkers/guardrails that block unauthorised actions (e.g. file deletion, external email)
  • an existing, widely-adopted AI assistant (Pair) to build the registry on top of

Common failure modes

  • agentic AI acting at machine speed without adequate human checks if registry coverage is incomplete
  • effectiveness unproven until the wider 2026 rollout is evaluated

Where it fits

Governance type
national whole-of-government digital agency
Scale
national — up to 150,000 public officers
Income level
high-income

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful