What if the real AI governance debate

isn't about how much regulation we need — but how we design it?

26 U.S. Attorneys General have called on Congress for federal oversight of advanced AI.

Common safety testing. Transparent incident reporting. Independent safety structures. But they added one crucial point:

States should still be allowed to introduce stronger safeguards.

This raises a governance question that feels surprisingly familiar from a Project Management perspective. Complex programmes rarely work through complete centralization or complete autonomy. They establish a common governance layer, while allowing individual workstreams to add controls when their specific risks require them.

AI governance could follow a similar architecture:

COMMON BASELINE → LOCAL CONTROLS → SHARED REPORTING → ESCALATION

Too much fragmentation creates complexity.

But complete standardization carries another risk: if the common framework misses something, everyone inherits the same weakness.

Perhaps the goal isn't to make every rule identical. It is to agree on what should never fall below the minimum.

Standardization doesn't necessarily require centralization.

So, how would you design AI governance?

One rule for everyone or one baseline with room for stronger safeguards?

Pubblicato originariamente su LinkedIn il 25/09/2026. Visualizza il post originale.