Regulation is loosening, not tightening
Through 2023 and 2024, governments seemed poised to put real brakes on AI development — the EU AI Act, the Biden executive order, the UK's Bletchley Park summit. Eighteen months later the direction has reversed. Major rules are being delayed, watered down, or repealed; safety institutes are being repositioned; and the political weather has shifted from "AI needs to be slowed down" to "AI must not be slowed down." The question is whether this is a temporary lull or a structural change in the politics of AI governance.
Timeline
- October 30, 2023
President Biden signs an executive order setting AI safety obligations across the US government. At the time, it is the most aggressive action any major government has taken on AI.
- August 1, 2024
The EU AI Act enters into force — the first comprehensive AI regulation by a major economy. Industry braces for compliance costs that some estimate at billions of euros.
- January 23, 2025
On his first day in office, President Trump signs an executive order revoking Biden's AI safety order and replacing it with a directive to remove regulatory barriers to AI development.
- April 9, 2025
The US AI Safety Institute is restructured into the Center for AI Standards and Innovation (CAISI), with a mandate focused on standards and competitiveness rather than safety testing.
- September 12, 2025
The European Commission opens consultation on an "omnibus simplification" of the AI Act — the first formal signal that the EU is reconsidering parts of its own regulation under industry and member-state pressure.
- May 12, 2026
The EU formally delays the high-risk-system provisions of the AI Act by 18 months, citing competitiveness concerns. The most demanding parts of the world's most advanced AI law are now scheduled to take effect in 2028, not 2026.
Where things stand right now
The political momentum on AI has flipped from "we must regulate this" to "we must not slow this down." Major rules are being delayed in the EU, repealed in the US, and softened almost everywhere they exist. Whether this lasts depends on whether AI produces a high-profile failure serious enough to swing public opinion back the other way — the structural pressure to deregulate is real, but a single major incident could reverse the direction quickly.