AI regulation is transitioning from policy to enforcement in 2026, with federal LLM procurement requirements landing in March, Colorado's algorithmic discrimination law in June, and EU high-risk AI obligations in August. Federal agencies now require model cards, evaluation artifacts, acceptable use policies, and feedback mechanisms. State laws focus on deployment harms like discrimination and transparency, while the federal government attempts preemption. Internationally, the EU delayed high-risk obligations to 2027, and China enforces content labeling and algorithm filing. Builders must prepare documentation, test deployed systems including tool use and agentic workflows, and create measurable, explainable AI behavior.
Table of contents
How Regulation Reaches Your ProductU.S. Federal PolicyEnforcement Without New LawsState LawsInternationalTechnical Context2026 TimelineWhat This Means for BuildersSort: