On the X axis, governance overhead per AI deployment. On the Y axis, deployment velocity. Across
ten jurisdictions a tradeoff is visible — but the fitted curve is not a straight line. There is a
sweet spot, somewhere in the low-50s, where governance signal is strong enough to be trusted but
not so heavy that it suffocates shipping. That is the curve to bend.
Composite scores synthesised from OECD AI Policy Observatory, Stanford AI Index 2025 and BCG Build
for the Future. Sixteen country composites plus four weighted regional averages (hollow rings). The
South America average is thinly observed and weighted heavily on Brazil. Direction-of-travel claims,
not precise estimates.
The thesis
Read the chart in three regions. Under-governed (X < 30): velocity is real but
unaccountable; trust collapses on the first incident. Sweet spot (X ≈ 45–58):
clear standards, lightweight assurance, defensible deployments. Over-governed (X >
65): procedural cost compounds; velocity halves. The most important policy job in 2026 is
not "more governance" or "less governance" — it is moving every jurisdiction toward the inflection.