AGI and Government
Public services, regulation, democratic process, and national security — how capable general AI affects how societies govern themselves.
Executive summary
Government has two roles in the AGI era: deploying AI to deliver better public services, and governing the development and use of AI itself. Both are now being attempted in earnest by major democracies, with the EU AI Act, U.S. federal frameworks, the UK AI Safety Institute, and the OECD AI Principles forming the backbone of the international regime.
Deploying AI in government
Tax administration, benefits processing, immigration triage, healthcare administration, and infrastructure maintenance are early adoption areas. Successes depend less on model capability than on data quality, procurement practice, and accountability for errors.
Governing AI
The EU AI Act (in force August 2024) created the first comprehensive legal regime, with specific obligations for general-purpose AI. The U.S. has used executive action, voluntary commitments, and sectoral regulators. The UK established the AI Safety Institute. The 2025 International AI Safety Report consolidated scientific evidence to inform policy.
Open questions
- How to evaluate frontier models before deployment.
- How to handle AI in democratic process (election content, voter manipulation).
- How to coordinate internationally on compute and capability thresholds.
- How to protect public-sector workers during transition.
Key takeaways
- 01Government is both user and regulator of AI.
- 02The international regime is taking shape rapidly.
- 03Quality of data and accountability matter more than model choice.
Frequently asked questions
Should governments use AI to decide benefits?
Cautiously and with strong appeal rights. Several countries (notably the Netherlands SyRI case) have seen high-profile failures.
Is regulation slowing innovation?
Evidence is mixed. The largest labs invest heavily in safety and welcome some regulation.