The case against age verification — even with zero-knowledge proofs

Position brief, 2026-06-07. Claims graded; empirical facts separated from argument. Formal core: models/z3/ageverif_futility.py. Companion: zkage/ (the strongest crypto case for age verification — and why even it fails).

Thesis: population-scale age verification / "age assurance" is futile-under-breach, a predator honeypot, adult surveillance by construction, and not rescued by zero-knowledge proofs. The correct posture is to oppose the requirement, not optimize the mechanism. Building a "privacy-preserving" version mainly legitimizes the mandate.

1. Futility under breach (formal — ageverif_futility.py)

A gate is effective only if credentials are unforgeable AND nontransferable. But:

• Every population-scale identity system is eventually breached → credentials become forgeable.
• Humans share and steal credentials → credentials are transferable (kids use adult creds;

adults use kid creds).

Therefore, post-breach, the system cannot distinguish authorized from unauthorized — it fails at its stated purpose. Z3 result: "mandate effective AND (forgeable OR transferable)" is UNSAT. The benefit is conditional and self-defeating; the surveillance infrastructure is unconditional. Once IDs are breached even once, IDs become meaningless for gatekeeping — anyone can be anyone. (Grade: STRONG — structural + empirical.)

2. The honeypot externality

Mandates create centralized linkage of real identity ↔ online behavior and databases of minors — precisely the target predators and hostile states want. The mandate manufactures the very vulnerability it claims to prevent, and by (1) it will leak. "Verify the kids" builds the kid-finder. (Grade: STRONG.)

3. Adult surveillance by construction

You cannot verify a child without processing every adult. The architecture surveils everyone; "protect the children" is the politically unassailable wrapper. This repo documents who actually benefits — Meta covertly funding the age-verification astroturf (influence-meta-childsafety), Oracle/Ellison→Tony Blair Institute (influence-tbi-policy), Altman/World ID (digitalid-corporate): the ID-surveillance complex, not children. (Grade: STRONG on architecture; beneficiary-steering documented.)

4. Zero-knowledge proofs do not save it

ZK improves the cryptographic disclosure of the input — not the systemic harms:

• (a) Predicate leakage: a proof of age≥18 still reveals that bit.
• (b) Presence/absence metadata: that you were asked, where and when, and whether you complied

is behavioral surveillance — independent of the cryptography.

• (c) Issuer centralization: a ZK credential still needs a trusted issuer = a centralized identity

root = the breach point and coercion lever from (1)/(2).

• (d) Normalization: it trains the credential-presentation reflex everywhere.
• (e) The legitimization trap: a "private" version is wielded to pass the mandate ("see, it can be

done safely"), accelerating adoption of the surveillance regime.

zkage/ demonstrates the best-case cryptography and its insufficiency. (Grade: STRONG.)

5. Accuracy harms (empirical)

Facial age estimation errs materially and unequally: Australia's own 2025 trial found ~±18 months error, worse for girls, First Nations, and low-SES users (digitalid-regulatory.json). Result: false gating and discriminatory exclusion layered on top of the surveillance. (Grade: STRONG, sourced.)

6. On intent (steelmanned, and why it doesn't matter)

The harms in (1)–(5) follow from the architecture, regardless of anyone's intent. So the case does not depend on proving a "deliberate dystopia" — that's the stronger argument: even granting good faith, the system surveils adults, endangers children, and fails at its goal. (Grade: intent not required.)

7. What to support instead (the alternative model)

• Device/OS-level, family-controlled tools — controls the user/parent owns, on the device, with

no platform or state identity layer.

• Default-private, minimize-by-design products; no population-scale identity prerequisite.
• Digital literacy / education; liability for demonstrated harms (the tort path, not the ID path).
• Oppose mandates in notice-and-comment rulemaking and the courts. It works: the UK "BritCard" digital-ID

compulsion was walked back after a 2.9M-signature petition (influence-tbi-policy).

Bottom line

Don't build a better age gate. Argue the category out of existence — it cannot be made safe, it cannot be made to work, and the "private" version is the trojan horse. Optimize opposition, not the lock.

Why the mandate keeps coming anyway (added 2026-06-11)

The futility case explains why age verification can't work; it doesn't explain why it keeps being mandated. That "why" is the control-rail incentive in digitalid-orchestration-real-incentive: age verification is the lever that de-anonymizes speech and money under an unbeatable child-protection frame — one track of the converging identity + programmable-money stack in digitalid-worldcoin-eid-convergence (Worldcoin → eIDAS wallets → Digital Euro → Chat Control's age-verification pivot → the UK Online Safety Act). Read together: this file shows the gate is futile and harmful; the orchestration block shows who benefits from building it anyway and why they can't say so.

Sources: EFF and privacy-research consensus on age verification; this repo's digitalid-regulatory.json, digitalid-corporate.json, influence-meta-childsafety.json, influence-tbi-policy.json, influence-ad-censorship.json; formal: models/z3/ageverif_futility.py.

← Research index · structured data: age-verification-abolition.json · age-verification-abolition.md