The honeymoon phase between the Trump administration and the titans of artificial intelligence is entering a period of cold, hard calculation. After a flurry of Day One executive orders aimed at stripping away what the President calls "ideological bias" and "bureaucratic friction," the White House is now quietly backpedaling on its promise of total deregulation. The shift is not driven by a sudden change of heart regarding safety or ethics, but by a chilling realization within the National Security Council: a completely ungoverned AI sector is a gift to state adversaries.
On January 20, 2025, the administration made good on a core campaign promise by rescinding the Biden-era Executive Order 14110. This effectively ended mandatory "red-teaming" and safety reporting requirements for frontier labs. However, by early 2026, the rhetoric has shifted from "unleashing" to "securing." The administration is currently debating a new framework that would mandate pre-release reviews for advanced models, particularly those that demonstrate "offensive cyber capabilities."
The Stargate Alliance and the Infrastructure Gamble
While the public focus has remained on executive orders, the real muscle of the current policy is a $500 billion infrastructure play codenamed "Stargate." This joint venture—spearheaded by SoftBank, OpenAI, and Oracle—aims to build the physical backbone of American AI dominance. It is an aggressive attempt to ensure that the hardware, data centers, and power grids required for the next generation of intelligence remain firmly on American soil.
But the Stargate strategy comes with a steep price tag for the tech giants. In March 2026, the administration forced a "Ratepayer Protection Pledge" on the industry's biggest players, including Microsoft, Google, Meta, and xAI. The deal is simple: if you want the government to clear the regulatory path for new data centers and nuclear reactors, you must pay for the infrastructure yourself. The companies are now legally bound to cover the costs of grid upgrades, ensuring that the massive energy hunger of AI does not drive up electricity bills for the average American household.
The Anthropic Conflict and the New Center for Standards
The administration's relationship with the industry is far from a monolith. While OpenAI has integrated itself into the federal strategy, Anthropic has become a lightning rod for tension. Following the company's refusal to grant the Pentagon unrestricted access to its "Mythos" model—a system capable of identifying thousands of high-severity software vulnerabilities—the administration designated Anthropic as a national security supply chain risk.
This "my-way-or-the-highway" approach to private tech has sent shockwaves through Silicon Valley. To manage this friction, the administration rebranded the NIST AI Safety Institute as the Center for AI Standards and Innovation (CAISI). Under the leadership of Commerce Secretary Howard Lutnick, CAISI has moved away from general safety "guardrails" and toward a strict focus on "demonstrable risks."
- Cybersecurity: Detecting models that can automate zero-day exploits.
- Biosecurity: Flagging capabilities related to the synthesis of chemical weapons.
- Interoperability: Establishing the technical protocols that will allow AI "agents" to communicate across different corporate and government platforms.
The Preemption War with the States
As the federal government pivots toward a more interventionist security posture, a second front has opened up in the legal system. California, Colorado, and Texas have spent the last year passing their own localized AI regulations, creating a "patchwork" that the White House views as a threat to national competitiveness.
In early 2026, the Attorney General launched an AI litigation task force specifically designed to sue states whose laws conflict with federal AI policy. The administration’s argument is that AI development is an "inherently interstate phenomenon" with foreign policy implications that supersede state-level police powers. This is not just a policy disagreement; it is a fundamental struggle over who owns the right to regulate the most powerful technology of the century.
The Jobs Paradox
The President recently conceded that while AI will "kill jobs," it will also "create a lot of jobs." It was a rare moment of nuance for an administration that usually leans into pure optimism. Behind the scenes, the Economic Policy Council is grappling with the reality that the productivity gains from AI may not distribute themselves evenly across the workforce. The current strategy relies on the "Winning the Race" action plan, which bets everything on a manufacturing and energy boom triggered by the need for AI hardware.
If the administration succeeds in building the largest fleet of power plants in U.S. history to fuel these models, they may indeed create a massive new blue-collar sector. If the models themselves then automate the very industries they were built to support, the administration will face a political crisis that no executive order can fix.
The current rethink in Washington is a transition from the "Wild West" era of deregulation to a "Fortress America" model. The goal is no longer just to let the industry run free; it is to ensure the industry serves as an auxiliary arm of national power.
