US, EU and Singapore take different paths on AI rules

As of October 2026, the United States, European Union and Singapore follow distinct approaches to AI oversight. The US has no comprehensive federal AI statute or enacted preemption of state laws, leaving obligations spread across state rules, sector regulators and existing laws. The EU uses the risk-tiered AI Act, while Singapore relies largely on voluntary frameworks, sector supervision and general laws, including the PDPA.

California’s frontier-model transparency law and several US chatbot laws are live, while Colorado’s replacement framework is reported for January 2027. In the EU, transparency duties apply and the AI Office can fine general-purpose model providers. High-risk obligations are scheduled for December 2027 and August 2028. Singapore has agentic guidance from IMDA and CSA in place, alongside final MAS AI guidelines. Dates reported by secondary sources require confirmation against official texts.

A shared governance foundation can address overlapping requirements: system inventories, risk tiers, documentation, logging, human oversight, adversarial testing and model-provider diligence. Companies can then add market-specific disclosures, US state notices and requests, EU classification and conformity requirements, and Singapore’s sector-specific expectations.