Unlike the European Union’s single, comprehensive AI Act, the United States has so far taken a considerably more fragmented approach to regulating artificial intelligence, with no single comprehensive federal AI law and a patchwork of different rules emerging from individual states, federal agencies applying existing laws to AI contexts, and an evolving federal policy stance that has shifted across different presidential administrations. Understanding this fragmented landscape requires looking at several different layers of regulation operating simultaneously, rather than expecting a single unified framework comparable to the EU’s approach.
Why the US Has No Single Comprehensive AI Law
Several structural and political factors explain why the United States has not followed the EU’s path toward a single comprehensive AI statute. The American federal system divides significant regulatory authority between the federal government and individual states, each of which can pass its own laws within its jurisdiction, and Congress has so far been unable to reach the kind of broad bipartisan agreement needed to pass comprehensive federal AI legislation, despite numerous bills being introduced and debated. Additionally, federal policy priorities around AI regulation have shifted considerably between different presidential administrations, with differing views on how much regulatory intervention is appropriate at this stage of the technology’s development, which has contributed to a federal approach characterized more by executive orders and agency guidance than binding, comprehensive legislation passed by Congress.
How Existing Federal Agencies Apply Current Law to AI
In the absence of AI-specific federal legislation, several existing federal agencies have asserted that current laws within their jurisdiction already apply to AI systems and have begun enforcing accordingly. The Federal Trade Commission has stated that existing consumer protection and anti-discrimination laws apply to AI systems making decisions about consumers, pursuing enforcement actions against companies for deceptive AI marketing claims or AI systems that produce discriminatory outcomes. The Equal Employment Opportunity Commission has issued guidance clarifying that AI-powered hiring tools remain subject to existing employment discrimination law, meaning an AI hiring algorithm that produces discriminatory outcomes can expose an employer to liability under laws that predate AI entirely. This approach of applying existing law to new AI contexts, rather than writing entirely new AI-specific rules, has become a significant feature of the current federal regulatory landscape.
California’s Leading Role in State-Level AI Regulation
California, home to many of the largest AI companies, has positioned itself as a leader in state-level AI legislation, passing multiple laws addressing different aspects of AI governance, including requirements around AI transparency and disclosure, rules specifically targeting deepfakes in political advertising and in non-consensual explicit imagery, and legislation addressing frontier AI model safety and risk reporting for the most capable models being developed. California’s outsized influence on AI regulation stems partly from its sheer market size and the concentration of AI companies headquartered there, which means compliance with California’s rules often becomes a practical necessity for AI companies regardless of where their other customers are located, echoing the kind of broad regulatory influence the EU exercises through its own market size.
Colorado’s Comprehensive AI Discrimination Law
Colorado passed one of the first comprehensive state laws specifically targeting algorithmic discrimination in high-risk AI systems, requiring companies that develop or deploy AI systems in consequential decisions, such as employment, housing, insurance, and financial services, to conduct impact assessments and take reasonable steps to prevent algorithmic discrimination based on protected characteristics. This law, which has faced amendments and delayed implementation timelines as businesses and lawmakers worked through practical compliance concerns, represents an approach more similar in spirit to the EU’s high-risk category system than most other state AI laws, focused specifically on consequential, rights-affecting AI applications rather than attempting to regulate AI broadly across all use cases.
New York’s Focus on Automated Employment Decisions
New York City implemented a specific local law requiring employers using automated employment decision tools in hiring to conduct independent bias audits of those tools and publicly disclose the results, along with notifying job candidates that an automated tool is being used in the hiring process. This narrower, sector-specific approach illustrates a pattern seen across several state and local AI laws: rather than attempting comprehensive regulation across every possible AI application, many jurisdictions have chosen to target specific, well-defined use cases where AI’s potential for harm, in this case employment discrimination, is particularly well understood and documented.
Other Notable State Approaches
Beyond these prominent examples, numerous other states have passed or proposed AI-related legislation addressing specific concerns, including laws requiring disclosure when AI is used in healthcare decisions, rules governing AI use in insurance underwriting, and various state-level data privacy laws that, while not AI-specific, significantly affect how companies can collect and use the data needed to train and deploy AI systems. This proliferation of different state approaches has created a genuinely complex compliance landscape for companies operating AI products nationally, since a single AI-powered product might need to comply with meaningfully different requirements depending on which state a given user happens to be located in when they interact with it.
The Ongoing Debate Over Federal Preemption
This growing patchwork of state laws has fueled an active policy debate over whether the federal government should establish a single national AI regulatory framework that would preempt, or override, the varying state-level rules, versus allowing states to continue serving as policy laboratories experimenting with different regulatory approaches. Proponents of federal preemption argue that a single national standard would reduce compliance complexity and provide greater legal certainty for businesses operating across state lines, while opponents argue that state-level experimentation allows regulatory approaches to adapt more quickly to a fast-evolving technology and prevents any single, potentially flawed federal approach from locking in rules that might prove inadequate or overly permissive. This debate remains genuinely unresolved, and the ultimate shape of comprehensive AI regulation in the United States will likely depend significantly on how this tension between state experimentation and federal uniformity plays out in the coming years.