Can regulation keep pace with AI's rapid evolution?

If generative AI's effect on inequality is deployment-contingent rather than predestined, then policy becomes the lever that decides which branch of the trade-off prevails. The review's diagnosis of current policy is unflattering: regulatory approaches in the European Union, United States, and United Kingdom sometimes fail to adequately address the challenges, because static, high-level guidelines lag behind the rapid advancement of the technology they aim to govern. By the time a rule is codified, the capability it targeted has shifted. The call is for a dynamic regulatory framework that can keep pace — one that maximizes AI's potential to reduce inequality while mitigating its harms.

The open question is what "dynamic regulation" concretely means and whether it is achievable. The pacing problem is structural, not incidental: legislative cycles measure in years, model releases in months, so any framework that specifies fixed thresholds or named capabilities is obsolete on arrival. The alternatives — outcome-based rules, adaptive standards bodies, mandated monitoring with revisable triggers — each trade legal certainty for responsiveness, and it is unresolved which trade is governable. This mirrors the human-centered-design problem from a different angle: there too, high-level guidelines failed to capture real-world nuance and lagged model evolution. The shared difficulty is governing a moving target without either freezing into irrelevance or dissolving into discretion. What process can steer deployment toward equality-reducing outcomes fast enough to matter is the question the review raises but cannot answer.