The definition · in four figures

How the parts connect

AI capabilities open possibilities. Human arrangements shape which possibilities are pursued, who benefits, and who can object. Repeated choices can become patterns; embedded patterns can become conditions that others inherit.

Four figures, each explained in words beside it, so you can follow the argument without the drawing.

Fig. 1 · What governance acts on

The system around the system

In wordsSocieties shape the arrangements around AI. Those arrangements shape what AI does in the world, from research to use. What it does becomes the conditions the next decisions start from.

Who decidesSocietiesGovernments, companies, researchers, workers, and the people affected.
What they shapeThe arrangements around AIIncentives · institutions · infrastructures · safeguards
What followsAI in the worldResearch, development, access, deployment, and use.

↺ And back again: today’s choices become tomorrow’s conditions

Fig. 2 · Four objects, pulling on each other

These dimensions interact

In wordsThe four objects are not separate boxes. An institution can change an incentive. Infrastructure can limit what an institution can inspect. A safeguard may depend on professional norms that encourage people to report problems. Power runs through every one of those relationships.

Institutions → IncentivesAn institution can change an incentive.
Infrastructures → InstitutionsInfrastructure can limit what an institution can inspect.
Incentives → SafeguardsA safeguard may depend on professional norms that encourage people to report problems.

Running through all fourPower: the ability to set terms, obtain information, refuse a deployment, demand an explanation, or secure a remedy.

Fig. 3 · Governance creates power too

What authority does a safeguard create?

In wordsA protective requirement can concentrate advantage. If independent evaluation is expensive or offered by only a few providers, established companies find it easier to comply and smaller organizations face new barriers. The answer is better design, not no design.

A ruleProtect peopleIndependent evaluation before certain systems can be deployed.
IfAccess is unevenEvaluation is expensive, or available from only a few providers.
ThenAdvantage concentratesEstablished companies comply easily; smaller organizations meet a barrier.
SoDesign for itProportionate requirements, credible evaluators, attention to who can access the process.

What power does this mechanism create—and who will hold it accountable?

Fig. 4 · Choices become conditions

Reversible in principle, expensive in practice

In wordsA decision that starts as a purchase can harden into a dependency. Nothing forbids reversing it, but staff, data and procedures have reorganized around it, so reversal becomes expensive and disruptive. That is why the definition says “as”: the change is already under way.

At firstA choiceWhether to buy an AI tool that helps clinicians identify urgent cases.Cost to reverse
ThenA patternStaff have been trained. Data formats have changed. Alternative procedures have disappeared.Cost to reverse
LaterA conditionWhether to renew a service around which hospitals have reorganized.Cost to reverse

The bars show the idea of rising cost, not measured data.

Do we have enough evidence to act?

What happens to our ability to act if we wait?

Four tests that run through every part

Whatever the object, ask the same four questions of it

PowerWho has the ability to shape the system, set terms, control infrastructure, or define the problem?
LegitimacyBy what authority are decisions made, who is represented, and who can contest them?
DistributionWho captures the benefits, who bears the costs and risks, and who is excluded?
AdaptabilityCan the governance mechanism learn, revise itself, and recover when its assumptions prove wrong?

Back to the definition, unpacked →Follow one decision through →