Trump Management Moves to Override State AI Laws, sparking Innovation vs. Regulation Debate
Table of Contents
The Biden administration is attempting to establish a national framework for artificial intelligence, potentially preempting a growing patchwork of state-level regulations that the White House argues are stifling innovation.
President Donald Trump signed an executive order on December 11, 2025, aiming to supersede state-level artificial intelligence laws deemed a hindrance to AI innovation.The move comes as states increasingly grapple with regulating the rapidly evolving technology, notably in the wake of powerful generative AI systems like ChatGPT.
Currently, 38 states have enacted laws in 2025 regulating AI in some capacity, ranging from prohibitions on stalking via AI-powered robots to restrictions on AI systems designed to manipulate human behavior. The administration’s order declares a policy of establishing a “minimally burdensome” national framework, signaling a preference for federal oversight.
According to the executive order, the U.S.attorney general will create an AI litigation task force to challenge state laws inconsistent with this new policy. Furthermore, the secretary of commerce is tasked with identifying “onerous” state AI laws and potentially withholding funding from the Broadband equity Access and Deployment Program for states enacting such legislation.Notably,state laws concerning child safety are exempt from this federal challenge.
“executive orders are directives to federal agencies on how to implement existing laws,” explained a senior legal analyst. “This order directs departments and agencies to take actions the administration believes fall within their legal authority.”
The push for federal preemption is not occurring in a vacuum. Big tech companies have actively lobbied for a unified national standard, arguing that navigating a complex web of state regulations is costly and hinders advancement.
California’s approach, for example, focuses on the most powerful AI models. The California law applies only to extremely large models, requiring at least US$100 million and 1026 – or 100,000,000,000,000,000,000,000 – floating point operations of computing power to train. As illustrated in a 2024 study by Rahman, Owen, and You, today’s most powerful AI models require significantly more computing power than their predecessors. [Image of graph showing floating point operations for AI models would be inserted here.]
These models, while powerful, are often described as “black boxes” due to their opaque internal workings. The risks associated with these large models include potential malicious use, malfunctions, and systemic risks, potentially leading to catastrophic consequences like weaponization or large-scale cyberattacks. The California law mandates developers to describe how they incorporate industry best practices and provide assessments of catastrophic risk, while also establishing a reporting mechanism for safety incidents.
Texas and Utah Prioritize Disclosure and Liability
Texas enacted the Texas Responsible AI Governance Act, imposing restrictions on AI systems used for behavioral manipulation.The law includes “safe harbor” provisions, offering liability protection to businesses that document compliance with responsible AI governance frameworks like the NIST AI Risk Management Framework. Uniquely, the Texas law establishes a “sandbox” – an isolated testing surroundings – for developers to safely assess AI system behavior.
The Utah Artificial Intelligence Policy Act focuses on disclosure, requiring organizations using generative AI tools to inform their customers.This law establishes that companies using generative AI are ultimately responsible for consumer liabilities and cannot attribute blame to the AI itself. It is the first state law to explicitly address consumer protections in the context of generative AI.
State Resistance and Broader Concerns
Despite the federal push, some states are resisting. Florida Republican Governor Ron DeSantis has voiced opposition to federal override efforts and proposed a Florida AI bill of rights. Moreover, the attorneys general of 38 states, along with those of the district of Columbia, Puerto Rico, American Samoa, and the U.S. Virgin islands, have collectively called on AI companies – including Anthropic, Apple, Google, Meta, Microsoft, OpenAI, Perplexity AI, and xAI – to address “sycophantic and delusional” outputs from generative AI systems.
The legality of the executive order remains uncertain, with observers suggesting that only Congress has the authority to supersede state laws. Though, the order’s final provision directs federal officials to propose legislation to achieve this goal, signaling a long-term strategy to establish federal dominance in the regulation of artificial intelligence.
