Governments converge on risk-based AI rules

AI governance is fragmenting across jurisdictions, but regulators are increasingly circling the same core ideas: risk-based controls, labelling and transparency, protections against discrimination, and human oversight for consequential decisions. The European Union has adopted the most comprehensive horizontal framework through the EU AI Act, while China applies detailed rules to generative AI services, content labelling, training data, and algorithm filing.

The United States remains a hybrid market, with a lighter federal posture and a growing patchwork of state laws covering automated decision tools, frontier models, training-data disclosures, watermarking, employment screening, and consumer interactions. The United Kingdom, Canada, Singapore, Japan, and Australia rely more heavily on principles, regulator guidance, voluntary standards, and existing privacy or sectoral laws rather than a single binding AI statute.

South Korea has enacted a comprehensive AI Basic Act, Brazil is advancing an EU-style risk-based bill, and Middle Eastern governments are centralizing policy through national strategies, ethics frameworks, and new AI authorities. Businesses operating across borders face a jurisdiction-by-jurisdiction compliance landscape where documentation, impact assessments, user notices, incident reporting, and human-review processes are becoming practical baseline expectations.