Study Warns the EU's AI "Guardrails" Are Too Rigid — a "Rigidity Trap" That Loses to the US "Leash" Approach
A Big Data & Society paper argues the EU's AI regime, forged through years of negotiation, is extremely hard to amend yet easy to repeal — a "rigidity trap" — and adapts to technological change worse than the US's on-demand "leash" model.
The "rigidity trap": hard to amend, easy to repeal
Per IThome (July 17), a paper in Big Data & Society warns the EU's AI "guardrail" mechanism falls short on both policy and enforcement and is too rigid to keep pace with fast-moving AI. Forged through years of negotiation and political coordination, it is extremely hard to amend yet relatively easy to partially repeal — a "rigidity trap." The EU AI Act, originally due in 2025, has now been replaced by an "AI Simplification Act."
Guardrails vs. a leash
The authors argue the current EU framework hasn't truly delivered on being "trustworthy, human-centric and respectful of fundamental rights." By contrast, US AI regulation is more like a "leash" — tightened as needed, more binding and faster to enforce than the EU's guardrails.
The trouble with predictive regulation
The core critique: the EU tries to predict in advance which risks AI will bring and build comprehensive rules, but no one can foresee every risk, making enforcement hard and limiting public value. The US typically legislates after a concrete risk appears, targeting a specific company or sector. The paper is a joint work of the University of Exeter (UK), the European University Institute and the University of Lausanne (Switzerland). For AI lawmaking rolling out worldwide, it raises a real question: against fast-evolving tech, is "comprehensive rules first" or "tighten as needed" more effective?