Verification tiers and proportionality

A forum account and a validator role on a financial network do not carry the same risk, so they should not demand the same evidence. That intuition — obvious in the physical world, where a bar checks ID but does not fingerprint you — is routinely ignored online, where services default to collecting the maximum because collecting is cheap and asking twice is annoying. The result is the familiar pathology: photocopies of passports held by businesses that needed to know one thing about you.

Proportionality is the design principle that fixes this: the intrusiveness of verification should match the risk of the action it protects. Tiered verification is proportionality made mechanical — a ladder of assurance levels, each backed by defined evidence, so that every service can ask for exactly the rung its risk requires and no more.

The wallet's four levels make the trade explicit. Unverified is the starting state — a wallet with no checks, fine for browsing. Basic reflects early checks: enough for low-risk activity where a service needs some confidence a real person is present. Standard — a verified government document plus a selfie under active liveness — is the workhorse level most services should target. Trusted adds hardware-backed biometrics, device integrity attestation and sustained history, and exists for genuinely sensitive roles such as validator onboarding.

Under the hood, the network computes a composite score from verification signals, and tiers are score bands — which means a tier is a summary the service can rely on without ever seeing the signals beneath it. The tier is the disclosure; the evidence stays put.

• For individuals: you invest effort and disclosure only when something actually warrants it. Nobody re-scans a passport to read a forum, and the decision to climb a tier is yours, made when a real action requires it.
• For services: asking for the minimum is the default path, because requests are structured around required tiers and specific proofs rather than free-form data collection. Over-asking becomes a visible, deliberate act — and one users can see and decline.
• For the network: eligibility checks are mechanical. An offering that requires Standard checks the tier; the marketplace approves or declines the action. Risk policy becomes code rather than judgement calls.

"Why not just verify everyone to the maximum once and be done?" Because evidence collected is risk created — for the person and the system — and because maximal verification imposes maximal friction on people who may lack the documents, the hardware or the ability to complete it. Proportionality is also an accessibility principle: every rung of assurance that risk does not require is a barrier someone did not have to face.

"Won't services just demand Trusted for everything anyway?" Some will try; structure pushes back. Requests display the demanded level to the user, who can see when a chat service demands validator-grade assurance and take their custom elsewhere. The guidance for services is blunt about this: over-asking erodes trust and adds nothing, and the network's own tier-gated actions model the right behaviour — higher requirements only where the stakes rise.

"Is a tier enough for a service to trust?" A tier is a compressed statement of evidence, published with its definition: what each level requires is documented, the scoring model is open source, and the result arrives cryptographically. That is a stronger basis than a service's own glance at an uploaded document — which, in the synthetic era, is no basis at all.

The deepest value of tiers is cultural: they keep the question "how much do we actually need to know?" alive in every integration, every design review, every request a user sees. Systems that stop asking that question drift toward maximum collection, because nothing in their structure resists it. A tiered, minimising, consent-bound architecture resists by default — and that, more than any single feature, is what makes an identity system trustworthy over decades rather than quarters.