Regulation of artificial intelligence in the United States
Based on Wikipedia: Regulation of artificial intelligence in the United States
On January 20, 2025, the regulatory landscape of American artificial intelligence fractured. In a single morning, President Trump signed an executive order revoking the landmark safety directives issued just sixteen months prior by his predecessor. The document, Executive Order 14179, did not merely tweak existing policy; it systematically dismantled the framework designed to enforce security standards on AI development. This was not a quiet administrative adjustment. It was a declaration that the federal government would no longer act as the primary gatekeeper of algorithmic safety, shifting the burden entirely onto the market and leaving a vacuum where oversight once stood. Just two months later, in March 2025, the administration began to articulate a new, aggressive stance: the patchwork of state-level AI laws was an impediment to national progress that required federal force to dismantle.
This rapid oscillation between stringent control and deregulation defines the modern era of American AI governance. It is a story not of steady evolution, but of violent pendulum swings driven by political ideology, technological anxiety, and the fierce competition for global dominance. To understand where the industry stands today, in mid-2026, one must trace the jagged path from the cautious optimism of the Obama years to the high-stakes volatility of the Trump-Biden-Trump cycle.
The roots of this regulatory struggle stretch back to 2016, a time when the existential threat of artificial intelligence was still largely confined to academic papers and science fiction novels. The Obama administration, recognizing the nascent power of machine learning, released a report titled "Preparing For the Future of Artificial Intelligence" in October of that year. The tone was measured, almost tentative. The National Science and Technology Council explicitly stated that the approach to regulation should be "informed by assessment of the aspects of risk," prioritizing the freedom of researchers to innovate over the imposition of heavy-handed constraints. It was a philosophy of "trust but verify," where the government acted as a facilitator rather than a warden.
That same year, Congress passed the John S. McCain National Defense Authorization Act, which included Section 1051. This legislation established the National Security Commission on Artificial Intelligence (NSCAI), an entity tasked with nothing less than comprehensively addressing the national security and defense needs of the United States in the age of machines. The commission was given a mandate to consider the methods necessary to advance AI development, effectively embedding the technology into the very DNA of American defense strategy. By 2018, the Artificial Intelligence Initiative Act was proposed, aiming to accelerate research for economic and national security purposes. The message from Washington was clear: the United States had to lead, but it had not yet decided how much to restrain its own children in the process.
The first real attempt at a unified regulatory philosophy arrived under President Trump's first term, albeit briefly. On January 7, 2019, following an executive order on maintaining American leadership, the Office of Science and Technology Policy (OSTP) released draft guidance containing ten principles for agencies. These were not laws, but they set the stage for a decade of debate: how do you regulate a technology that moves faster than legislation? The National Institute of Standards and Technology (NIST) responded with position papers, and the Defense Innovation Board weighed in on ethics. Yet, the approach remained fragmented. Agencies like the Food and Drug Administration began carving out their own paths, creating specific pathways to regulate AI in medical imaging, while the broader picture remained a collection of disjointed efforts.
The turning point came with the release of the NSCAI's final report in March 2021. The tone had shifted dramatically from the cautious optimism of 2016. The report warned that advances in AI could lead to "inflection points or leaps in capabilities" that might introduce new concerns and risks requiring entirely new policies. It spoke of systems needing to be aligned with goals and values, including safety, robustness, and trustworthiness. This was no longer about facilitating research; it was about survival. In June 2022, Senators Rob Portman and Gary Peters introduced the Global Catastrophic Risk Management Act, a bipartisan effort explicitly designed to counter the risk of AI being abused in ways that could pose a catastrophic threat.
Then came October 4, 2022. President Joe Biden unveiled the AI Bill of Rights, a non-binding but morally weighty document outlining five protections for Americans: safe and effective systems, protection against algorithmic discrimination, data privacy, notice and explanation, and human alternatives or fallbacks. It was a statement of values, a declaration that in the AI age, citizens should not be mere subjects of black-box algorithms. The following year, the administration moved from principles to pressure. In July 2023, seven tech giants—Amazon, Anthropic, Google, Inflection, Meta, Microsoft, and OpenAI—signed voluntary commitments to manage risks. They promised internal and external security testing, watermarking for AI-generated content, and a focus on societal challenges like cancer prevention and climate change. Eight more companies, including Adobe and Nvidia, joined later that September.
But voluntary commitments were fragile things. The administration knew it needed the force of law. On October 30, 2023, President Biden issued Executive Order 14110. This was the most comprehensive federal action on AI to date. It directed agencies to develop standards for critical infrastructure, mandated rigorous cybersecurity testing for dual-use models, and leveraged the federal government's massive purchasing power to shape the market. It was a bold assertion that the state had a right—and a duty—to police the development of powerful algorithms.
That order stood for less than two years. The political winds shifted with the 2024 election results, and in January 2025, President Trump revoked Executive Order 14110. In its place came Executive Order 14179, which stripped away many of the safety mandates. The administration argued that these regulations were stifling innovation and handing the lead to China. But the deregulation did not stop at the federal level; it sparked a war with the states.
By December 2025, the conflict had escalated into a full-blown legal confrontation. President Trump signed Executive Order 14365, an order that fundamentally altered the relationship between Washington and the state capitals. It directed federal agencies to develop a unified national approach to AI policy while simultaneously evaluating state AI laws for potential conflicts. The order did not merely suggest that states align with federal priorities; it threatened legal action against any state that passed regulations contradicting the new federal stance. Perhaps most significantly, it conditioned certain federal funding on state compliance, effectively holding grants and infrastructure dollars hostage to force alignment.
The exceptions were telling: state laws related to child safety, data center infrastructure, and state government procurement were exempted. This carve-out revealed a complex reality. The administration was willing to allow states to protect children and build physical infrastructure, but it drew the line at regulating the behavior of AI companies or the content they generated. The message was clear: the federal government would be the sole arbiter of how artificial intelligence is developed and deployed.
This centralization clashed violently with the reality on the ground. While the Trump administration criticized state lawmakers and threatened lawsuits, states continued to act. California, a global hub for technology, had already enacted laws in 2025 that went beyond federal requirements. Other states were poised to pass legislation in 2026 and 2027. The administration viewed this "patchwork" of regulations as an existential threat to the industry's efficiency. They argued that a company should not have to navigate fifty different sets of rules for five hundred million people. Yet, state legislators saw their own responsibilities to their constituents, fearing that without local protections, Americans would be exposed to discrimination, privacy violations, and unsafe systems.
The TAKE IT DOWN Act, passed by Congress in 2025, represented a narrow victory for the bipartisan desire to curb specific harms. Targeting AI-generated deepfakes, particularly those used for non-consensual pornography or election interference, it was one of the few areas where federal consensus held. But even this success highlighted the broader paralysis. The law addressed symptoms—deepfakes—while the underlying regulatory framework remained in flux.
The debate over regulation has always been a battle between two competing imperatives: innovation and safety. The Obama administration prioritized the former, fearing that premature regulation would choke off American ingenuity. The Biden administration attempted to balance both, introducing guardrails while encouraging growth. The Trump administration of 2025-2026 has swung hard toward unbridled innovation, arguing that the only true risk is falling behind in a global arms race. But this framing ignores the human cost of unchecked algorithms.
When an algorithm denies a loan based on flawed data, or when a deepfake sways an election, the abstract concept of "regulation" becomes immediate and personal. The AI Bill of Rights, though non-binding, articulated what was at stake: the right to know why an algorithm made a decision, the right to be treated fairly, the right to human oversight. These are not technical specifications; they are civil liberties. When the federal government steps back from enforcing these rights, as it did in 2025, states often step forward to fill the void. This is not merely bureaucratic friction; it is a struggle over who gets to define the future of American society.
The executive order of December 2025 attempted to solve this by imposing a single federal vision. It directed agencies to challenge state laws in court and to use funding as leverage. But history suggests that such top-down mandates often fail to address the nuances of local reality. A regulation that makes sense for Silicon Valley may be disastrous for a rural community concerned about job displacement or data privacy. The administration's strategy assumes a uniformity of risk that does not exist.
Furthermore, the speed of technological change renders static laws obsolete almost before they are signed. The NIST AI Risk Management Framework, released in January 2023 as voluntary guidance, was designed to be adaptable. But when the government shifts from voluntary frameworks to hard-line executive orders and back again within a span of three years, it creates uncertainty that can be just as damaging as over-regulation. Companies cannot build long-term strategies if the rules change with every election cycle.
The role of the courts has become increasingly pivotal in this drama. As states pass laws to protect their citizens and the federal government moves to strike them down, the judiciary will be forced to decide the scope of state power in the digital age. Will the Constitution's commerce clause allow a state to regulate an AI model developed in another country but used by its residents? Can the federal government use funding threats to coerce states into abandoning their safety standards? These questions are no longer theoretical; they are being litigated right now.
The human cost of this regulatory limbo is measured in lost jobs, biased hiring decisions, and eroded trust. When a hospital uses an AI system trained on biased data to triage patients, the consequences are not abstract statistics. They are people denied care, families torn apart, lives shortened. The Obama administration's 2016 report noted that regulation should be informed by risk assessment. But who defines the risk? Is it the speed of innovation? Or is it the safety of the patient?
The Biden administration's AI Bill of Rights attempted to answer this by centering the human experience. It demanded notice and explanation, insisting that people have a right to understand the automated decisions that govern their lives. The voluntary commitments signed in 2023 promised watermarking and transparency, acknowledging that users deserve to know when they are interacting with a machine. These were steps toward a more humane digital future.
Yet, the reversal of these policies in 2025 suggests a different priority: speed above all else. The argument is that the United States must move faster than its competitors, regardless of the regulatory friction. But this logic ignores the fact that trust is a prerequisite for adoption. If Americans do not believe their data is safe, if they fear their voices are being manipulated by deepfakes, they will reject the technology regardless of how fast it develops.
The state of AI regulation in 2026 is thus one of profound contradiction. Federal policy oscillates between strict control and radical deregulation. States fight to maintain their autonomy. Companies navigate a maze of conflicting requirements. The public waits for clarity that never comes. This instability is not just a regulatory failure; it is a reflection of a deeper societal anxiety about the future.
The events of 2025 and 2026 have shown that the path forward is not linear. It is a jagged line drawn by political will, technological capability, and human courage. The executive orders, the lawsuits, the state laws—these are all battles in a war for the soul of artificial intelligence. Will it be a tool for empowerment and safety? Or will it become an instrument of control and chaos?
The answer lies not just in Washington or Silicon Valley, but in the courts, the states, and the public sphere. It lies in the willingness of lawmakers to prioritize human dignity over corporate speed. It lies in the ability of society to demand accountability from machines that are increasingly making decisions about our lives.
As we look toward 2027 and beyond, the stakes have never been higher. The regulatory framework that emerges from this chaos will define the relationship between humanity and its most powerful creation for generations. The patchwork of state laws, the federal mandates, the voluntary commitments—these are all threads in a tapestry that is still being woven. And the pattern it forms will determine whether AI serves us or rules us.
The story of American AI regulation is not over. It is entering its most critical chapter. The decisions made in courtrooms and legislatures today will echo for decades. The question is no longer whether we can regulate AI, but how quickly we can do so before the consequences become irreversible. The clock is ticking, and the future is being written right now.