The pace of AI regulation has accelerated. What began as a handful of regional frameworks is now a patchwork of binding laws, voluntary codes, and enforcement activity. For global AI leaders, the challenge is no longer “if” compliance will be required — it is aligning operations so that models, products, and governance can flex across borders without slowing delivery.

Here’s where the world’s major AI jurisdictions stand, what’s coming next, and the operational steps to take before Q1 2026.

European Union

Status and near-term obligations
The EU AI Act is now in force, with staged applicability over the next three years. Prohibitions on unacceptable-risk systems are already live. Governance requirements, GPAI model obligations, and conformity assessments for high-risk AI will phase in through 2026 and 2027. Providers and deployers must also prepare for new oversight from market surveillance authorities.

What high-risk providers and deployers must prepare
Expect obligations around post-market monitoring, incident reporting, technical documentation, human oversight, and formal conformity steps. These are not one-time exercises — they will need continuous upkeep.

Practical actions
Conduct a gap assessment against high-risk requirements now. Assign clear ownership for technical documentation. Build a process to manage regulator interactions, including market surveillance requests.

United States

Federal posture
The White House AI Action Plan outlines more than 90 actions spanning innovation incentives, infrastructure investment, and international coordination. While there is no single federal AI law, momentum is building for sector-specific guardrails.

Enforcement vectors
Agencies like the FTC and CFPB are using existing laws to target deceptive, unsafe, or discriminatory AI practices. This means enforcement can hit before new AI-specific rules are passed.

State dynamics
State approaches diverge sharply, from California’s detailed AI risk assessments to more procurement-driven policies in others. For multi-state operations, this creates a moving target for compliance.

Practical actions
Map each AI use case to applicable consumer protection and sector rules. Track federal grants and infrastructure incentives that could offset compliance costs.

United Kingdom

Direction
The UK continues a pro-innovation, regulator-led approach. The AI Safety Institute has taken a central role in model evaluation, and moves are underway toward more binding arrangements.

Practical actions
Engage early with regulator guidance. Where relevant, align your evaluation and testing protocols with the Safety Institute’s frameworks to smooth future certification or procurement.

BRICS and the Global South

Rio Declaration themes
BRICS leaders are calling for inclusive AI governance, sustainable development alignment, and coordination through the UN. This reflects a push toward multilateral AI principles rather than region-specific rulebooks.

Practical actions
Adopt a governance model that can flex to different cultural values and local data rules. This is critical if operating in markets where AI trust frameworks differ significantly from Western approaches.

Canada and Singapore

Canada — AIDA status
The Artificial Intelligence and Data Act (AIDA) is moving toward implementation, using a scope- and risk-based approach. Companion documents are clarifying definitions and operational expectations for high-impact systems.

Singapore — Model AI Governance for GenAI
Singapore’s updated framework remains one of the most pragmatic global references. The 2025 refresh includes sector-specific guidance and examples for generative AI use cases, offering a playbook that balances trust with innovation.

Building a cross-border operating model

Global AI operations demand consistency without losing local relevance. The most efficient approach is to keep a central policy with local addenda for each jurisdiction.

Collect evidence once, reuse it many times for audits, tenders, and regulator requests. Tag AI products by jurisdiction at launch so requirements are tracked from day one. Include contract clauses covering model transparency, incident response, and data provenance.

Your 60-day action plan

• Assign a global AI policy owner with authority to coordinate across legal, product, and engineering teams

• Build a jurisdiction matrix mapping AI rules and overlay relevant controls

• Prepare a documentation bundle that can be repurposed for audits, tenders, and regulator inquiries

By Q1 2026, AI governance will no longer be optional in any major market. The organisations that win will be those who build governance into their operational DNA, rather than bolting it on when regulators arrive.