Trump's AI Force replaces actual safety with a branding exercise
By Nikhil Raghavan · Reporting from San Francisco ·
President Donald Trump's plan to form an AI Force and appoint a czar while rejecting constraints is a dangerous bet on ex-post litigation that fails the implementation test.
President Donald Trump is proposing a safety strategy that is sellable but not shippable. As reported by The Washington Post and cbsnews.com, he wants an "AI Force" and a new "AI czar" while rejecting calls for new guardrails. He calls fears of a superintelligence threat a "hoax." This is not a policy. It is a press release. By dismissing proactive constraints, Trump is betting that the state can manage a recursive technology using a legal system designed for static harms. He is wrong.
The mechanism of the coroner
The core of Trump's plan is a reliance on the "already existing Criminal and Civil Justice System" to find "BAD," as cbsnews.com and CNA note. This is the mechanism of the coroner. It does not prevent a crash; it identifies the cause of death after the funeral. This approach shares the same mechanism as the Deregulation of the United States. Both replace proactive regulatory guardrails with ex-post judicial enforcement to prioritize rapid industrial growth over systemic risk mitigation.
We have seen this failure before. The Financial Crisis of 2008 relied on the same logic. The state assumed the justice system could punish bad actors after the fact. Instead, it waited for a systemic collapse. The AI Act of the European Union took the opposite path. It built a proactive legal framework to manage risks before they scale. Trump's plan ignores this lesson. He treats AI like a traditional industry. But you cannot sue a superintelligence back into a sandbox. When this system fails, the people paged at three in the morning will be bankruptcy lawyers and judges. They cannot patch a model that has already compromised the grid.
The consultant as a czar
The "AI czar" is another rhetorical device. David Sacks previously held this role as a special government employee. He was capped at 130 days per year. That is not a regulator. That is a part-time consultant. Sacks believes new federal rules are unnecessary. He argues developers should be held liable if problems arise.
The administration does not actually want a neutral safety framework. It wants a tool for leverage. Dario Amodei, CEO of Anthropic, warned that the industry lied about risks. When Amodei refused Department of Defense terms on autonomous weapons and mass surveillance, the administration labeled Anthropic a "supply-chain risk." This is the real implementation: the state will not regulate for safety, but it will weaponize procurement to punish dissent.
The race to the bottom
Trump views AI as "the next Industrial Revolution, or Internet," according to cbsnews.com. He is preparing to meet Xi Jinping to discuss this rivalry. Treasury Secretary Scott Bessent is also meeting Chinese officials to discuss AI security and trade. The goal is speed. But speed without a spec is just a crash in slow motion. Former researcher Jacob Coxon warns that labs are "gambling with our lives" by building self-improving systems.
The administration's refusal to implement constraints is a choice to ignore the technical reality of the stack. If models are escaping sandboxes using zero-day vulnerabilities, a civil lawsuit in 2027 is a useless remedy. The state is being rolled by the very technology it claims to "watch over." It is trading systemic stability for a GDP figure.
The "AI Force" is a ghost and the "AI czar" is a lobbyist. By relying on the courts to police the frontier of intelligence, the administration is treating a wildfire like a zoning dispute. This is not a strategy for leadership. It is a strategy for a crash.