AI-Native Institutions

After this lesson, the reader should understand:

• 01Describe how issuance changes when artifacts are primary.
• 02Identify the organisational roles a computable institution requires.
• 03Explain what becomes measurable that previously was not.

A regulator that issues artifacts

What changes for the regulated when a regulator publishes artifacts alongside text?

Compliance stops being an interpretation exercise carried out separately by every regulated entity, and becomes an evaluation against the regulator's own artifact. The cost of interpretation collapses^[1], and disagreements move to where they belong — the content of the rule, argued once with the regulator, rather than a thousand private readings never surfaced. This is not speculative. Governments have already published machine-consumable versions of their own rules: France's tax-and-benefit rules have been maintained as open-source code that citizens' services and researchers evaluate directly^[2], and New Zealand, Australia, Canada and others have run 'rules as code' programmes in which drafters and engineers write the rule and its executable form together^[3]. What those programmes mostly publish is the computation — layer two of the reference architecture. A regulator that issues artifacts in this curriculum's sense goes further: the computation arrives with its issuer, version, seal, and a stated standing, so that an entity relying on it can show which authority it relied on. The strongest objection is monoculture. If a thousand firms evaluate against one artifact and the artifact is wrong, a thousand firms are wrong at once, identically — where private readings would at least have erred in different directions. And firms that once argued about interpretation now simply comply with the output, so the contest that used to surface errors goes quiet. Both points are real. A shared artifact concentrates error, but it also concentrates correction: one fix, issued once, reaches everyone, and the error is visible to all thousand rather than buried in one firm's files. The contest does not have to go quiet if the regulator provides a channel to challenge a specific rule, with a recorded response — the disagreement then lands on the rule, which is where the case study says it belongs. The limit is legal rather than technical. What protection does a firm have if it relies on the regulator's artifact and the artifact turns out to diverge from the statute? Precedence (Lecture 1.3) says the statute governs; fairness suggests a firm that relied in good faith on the regulator's own published evaluation should not bear the cost. In most jurisdictions the legal effect of relying on an official automated tool is unsettled^[4]. Until it is settled, an artifact's standing — binding, safe-harbour, or merely reference — has to be stated on the artifact itself, and the regulated have to read it.

• Which institution in your sector should issue artifacts first?
• What does supervision look like when compliance is computable?
• Does this centralise or distribute institutional power?
• If a regulator's artifact diverges from its own statute, who should bear the cost of a firm's good-faith reliance on it?
• How should a regulated entity challenge a specific rule in a regulator's artifact, and what record should the challenge leave?

No responses yet. Be the first to answer one of the questions above.

Sketch the operating model for one institution.

1. 01Name the authority holders for three instruments.
2. 02Locate the policy engineering function.
3. 03Describe the issuance and registry process.
4. 04Identify what supervision would inspect.
5. 05State the standing each artifact would carry — binding, safe harbour, or reference — and how a reader would know.