AGI Policy and Regulation in 2026
The state of the global regulatory architecture for AI — what is in force, what is coming, and where the open debates are.
Executive summary
The global regulatory architecture for AI now includes the EU AI Act, the NIST AI RMF, the UK and US AI Safety Institutes, OECD AI Principles, the International AI Safety Report, and a growing set of bilateral agreements. The pattern is risk-tiered: lightest regulation for low-risk uses, strongest oversight for frontier models and high-risk applications.
Key concepts
- Risk-tiered regulation
- Frontier oversight
- Sectoral application
- International coordination
- Open debates
What is in force
EU AI Act (phased through 2026), NIST AI RMF (voluntary in the US), national AI strategies in most G20 countries, and active oversight institutes in the UK, US, EU, and several others.
What is coming
Frontier-model registration and evaluation requirements, mandatory incident reporting, sector-specific regulation in health and finance, and international coordination mechanisms.
Open debates
Compute-based regulation thresholds, liability frameworks, open-weight model treatment, content provenance, and the global governance of frontier training.
International coordination
G7 Hiroshima process, AI Safety Summit follow-ups, OECD and UN frameworks, and bilateral US–EU and US–UK agreements.
Key takeaways
- 01Risk-tiered regulation is the dominant frame.
- 02EU AI Act is the most comprehensive framework in force.
- 03Frontier-model oversight is the most contested area.
- 04International coordination is real but uneven.
Frequently asked questions
Does the EU AI Act apply to me?
If you provide AI systems used in the EU, very possibly. Compliance starts with classifying your system's risk tier.
Will there be a global AI regulator?
Unlikely in the near term. Coordination through OECD, UN, and bilateral agreements is the more realistic path.
Further reading
Related hubs
How AGI is reshaping policy, public services, democratic process, and geopolitics — and what citizens should expect.
How AI is being deployed in benefits, tax, immigration, healthcare, and other public services — and the trust, audit, and procurement requirements.
How AI-generated content, election integrity, and information environments are being affected — and the countermeasures being built.