How Trump Was Persuaded to Regulate A.I.

What It Covers

NYT reporter Tripp Mickle traces how Trump signed an executive order requiring AI companies to voluntarily share models with the government up to 30 days before public release — a reversal driven by cybersecurity fears, internal White House conflict between David Sacks and Scott Bessent, and pressure from JPMorgan's Jamie Dimon and Microsoft.

Key Questions Answered

• •Catalyst for regulation: Anthropic's unreleased model, Mythos, demonstrated the ability to identify software vulnerabilities at scale, alarming Microsoft and JPMorgan CEO Jamie Dimon enough to call the White House directly. This single model's capabilities shifted the administration from zero oversight to drafting an executive order within weeks, illustrating how one concrete threat can override ideological resistance.
• •Internal White House fault lines: Treasury Secretary Scott Bessent and Chief of Staff Susie Wiles drove the regulation push, fearing political liability if a cyberattack occurred under a hands-off policy. AI czar David Sacks opposed it entirely. Understanding which officials hold economic versus innovation portfolios reveals where future AI policy battles will actually be decided inside this administration.
• •How the order was weakened: The review window shrank from 90 to 30 days after David Sacks personally called Trump the morning of the original signing ceremony. A clause was added explicitly prohibiting mandatory government licensing or preclearance. Tracking these specific textual changes shows how industry-aligned advisors can reshape policy language in hours, not months.
• •Populist opposition from both flanks: Steve Bannon frames AI as incompatible with a free society and has mobilized over three dozen pastors who cite AI companion relationships destroying marriages. Bernie Sanders proposes a 50% government ownership stake in major AI companies via an AI Sovereign Wealth Fund at IPO. Both factions want mandatory oversight, not voluntary disclosure.
• •Regulation may require a crisis first: AI currently accounts for one-third to more than half of US GDP growth, giving the industry enormous political leverage. Social media went 15 years through repeated crises — polarization, youth mental health damage, child exploitation material — without a single binding rule. AI regulation may follow the same pattern unless economic pain spreads beyond current concentrated beneficiaries.