Regulation, Law & Liability
Limited binding AI regulation. Rapid innovation, but open questions on liability, safety and the concentration of power.
Where does the USA stand, and where does it break?
Unlike the EU, the United States has no comprehensive federal data protection law and no unified AI legislation. Instead, regulation is federally fragmented: in the absence of federal legislation, several states have enacted their own rules. California introduced the CCPA/CPRA for data protection; Colorado followed in 2024 with the first comprehensive AI law enacted by a US state, since revised on multiple occasions (Colorado AI Act). This fragmentation creates a patchwork of varying obligations across states: companies face a mosaic of partly contradictory rules, while at the federal level neither an AI law nor a general data-protection statute exists. Sectoral regulators such as the FTC, the SEC and antitrust authorities apply existing powers to individual cases, but a coherent framework is lacking. Who is liable when an AI system makes a faulty decision, discriminates or causes harm? The question remains largely unresolved in law and is increasingly being contested in the courts and through state tort law. Federal policy oscillates between promoting innovation and targeted oversight, shifting with the political majority.
In the economy ahead, AI systems make decisions about credit, hiring, medical recommendations and administrative acts. This raises the question of who is liable for errors, discrimination and harm. The United States risks technology outrunning regulation. Without a federal framework, liability remains fragmented and unresolved, while a small number of corporations are consolidating — through foundation models, computing power and data — a degree of market power that existing competition and liability law struggles to address. The state-level patchwork impedes consistent standards and creates legal uncertainty for both those affected and businesses. Innovation-friendly regulation secures pace and leadership, yet without clear liability rules and guardrails, harm, discrimination, and a concentration of power that is difficult to hold democratically accountable all become real risks. Whether the United States can establish a coherent framework that enables innovation while securing liability, safety, and competition will determine the trust and balance of power in the AI economy.
Paths for the USA
The United States enacts a federal data-protection and AI law that replaces the state-level patchwork with uniform standards. This strikes at the heart of the structural fault line, because fragmentation creates legal uncertainty and unified liability rules are absent. A federal framework would give both businesses and those affected the clarity they need. The constraint lies in political deadlock: there are deep partisan divisions over scope, federal pre-emption of state law, and enforcement — divisions that have blocked a compromise for years.
Trade-off: legal clarity versus federal and partisan conflict
Feasibility: Mooted for years, but unable to secure a majority
The United States establishes clear rules on who is liable when AI systems make faulty decisions, discriminate or cause harm — through legislation or the further development of liability law. This matters because unresolved liability leaves those affected unprotected and businesses in the dark. US product liability and tort law offer points of reference. The limitation: assigning liability for complex, self-learning systems is legally difficult, and a patchwork of court rulings risks emerging in place of a coherent framework.
Trade-off: protection and clarity versus legal complexity
Feasibility: Developing through the courts rather than legislation, for now
The United States deliberately leverages federalism: states such as California and Colorado test AI and data rules, with the most promising approaches potentially consolidating into federal standards over time. This fits the federal tradition and permits experimentation without nationwide risk. Successful models can serve as templates. The downside is precisely the fragmentation: as long as no federal framework follows, the patchwork persists, and businesses bear the costs of contradictory obligations.
Trade-off: freedom to innovate versus uniform standards
Feasibility: Only valuable if insights from state-level experiments are consolidated at the federal level
The United States is sharpening antitrust and competition law to prevent a handful of corporations from consolidating market power over foundation models, compute and data in ways that are difficult to govern. This addresses the concentration of power that existing law struggles to capture. The FTC and DOJ have initiated several proceedings and are examining AI-specific questions. The limits are clear: cases take years, the economics of network effects and economies of scale structurally favour concentration, and intervention risks slowing innovation.
Trade-off: open competition versus loss of innovation and pace
Feasibility: Effective only if the law reimagines the AI economy
Trigger-based phases
Trigger: AI systems spread through decision-making processes while neither data protection nor AI legislation exists at the federal level.
States fill the gap with their own rules, creating a patchwork. Liability questions remain unresolved and migrate to the courts. Legal uncertainty grows for those affected and for businesses alike.
Trigger: The number of state-level rules grows, and pressure for a unified federal framework intensifies.
Businesses bear the costs of conflicting obligations; those affected face highly uneven levels of protection. The dispute over a federal framework — its scope, preemption and enforcement — will determine whether rules converge or continue to diverge.
Trigger: It becomes clear whether the United States will find a coherent framework for data protection, AI and liability — or entrench the fragmentation.
If a federal framework is achieved, trust and legal certainty will follow. If the patchwork persists, liability and the concentration of power will remain unresolved. The choices made in these years will shape the governance of the AI economy through to 2040.
Real-world pilot programmes
Status: Enacted, revised several times, entry into force deferred to 2026 (first comprehensive AI law passed by a US state)
Colorado became the first state to regulate the use of AI in high-stakes decisions, requiring due care against algorithmic discrimination. As a laboratory of federalism, the law illustrates how AI regulation takes shape in the United States — and how fiercely contested its scope and implementation are.
Status: In Kraft (umfassendster Datenschutz eines US-Bundesstaats)
In the absence of federal legislation, California's data protection law de facto sets the standard for many US companies. It is the test case for whether state-level rules can fill a protection gap — and simultaneously makes visible the fragmentation that persists without a federal framework.
Status: In force (voluntary framework for managing AI risks)
The NIST framework offers companies a voluntary standard for assessing and mitigating AI risks. As a non-binding reference framework, it illustrates the innovation-friendly US approach — guidance rather than prescription — and its limitations with respect to liability and enforcement.
What we are watching
This indicator measures whether a coherent federal framework is emerging. It matters because a growing number of diverging state laws is intensifying legal uncertainty. Without a federal framework, the patchwork will harden. Trust and investment security require a coherent framework.
This indicator measures how clearly it is established who is liable for AI-related harm. It matters because unresolved liability leaves those affected unprotected and companies operating in uncertainty. If it remains open, risk grows with every automated decision. Legal protection requires clarity.
This indicator measures the degree to which model, compute, and data power is concentrated among a handful of providers. It matters because concentration without a counterweight becomes difficult to govern democratically. If left unchecked, power will shift permanently to a small number of corporations. This indicator is central to both competition and governance.