The International Foundation for Child-Safe internet Standards develops the metadata standard and admission criteria for a governed namespace — a space designed so that harmful material is absent because it was never admitted, rather than removed after it appears.
Filtering, blocking, moderation and takedown all act after publication — after a child has already been reached. They are treatments applied downstream of a supply that is universal, unlicensed and unbounded.
An admission-based namespace works differently. Publishers apply; published criteria are applied before an address is issued. Admission is a tenancy: accept the terms or do not enter, and breach the terms and lose the address. Nothing is inspected in transit, because nothing needs to be.
One canonical, machine-readable metadata schema, published openly and verifiable by anyone.
The format is invariant everywhere; the thresholds inside it are set by each jurisdiction.
A household election, never a mandate. The open internet continues unchanged alongside it.
Land-use law has governed what may be built where for a century, through criteria published in advance and applied at the point of permission. The addressing layer works the same way: addresses are issued under agreement, conditioned, recorded, and withdrawn when the terms are breached.
How the standard works →Children have always been protected in part by geography — libraries, schools, cinemas and shops have physical edges, and those edges did real protective work. Point-to-point addressing has no edges: schoolwork and a self-harm forum sit the same distance apart, one reach. Exposure no longer requires seeking; it requires only adjacency.
The prevailing model treats the internet as a river to be treated downstream — but the upstream is universal, there is no chokepoint, and treatment always acts after exposure. A block list is a record of harms already discovered.
02 · Admission, not filteringA governed namespace is additive, not subtractive: it is built by admitting vetted material into a bounded space, not by filtering harmful material out of an open one. Nothing is inspected in transit and no traffic is intercepted — the architecture operates on a different layer entirely from the one encryption protects. That produces a defined, verifiable boundary other systems can bind to.
What follows describes a model, not a service in operation.
Governs who enters. A publisher applies, identity and purpose are verified, published criteria are applied, and a registry agreement is signed as a condition of the address.
Governs who remains. Content changes after admission, so review is continuous — failure returns the registrant to the admission question. Either gate alone is a label.
Enforcement runs in two directions: vertically, registry to registrant — admission required accepting the terms, so breach permits rapid suspension without litigation; and horizontally, registrant to registrant — every participant enters on the same published terms and so has standing against the others.
The standard governs what is inside the district — it does not claim to prevent a determined adolescent from leaving it. A child using a VPN is outside the namespace, not leaking through it. That is a coverage boundary, not a system failure.
An architecture that can admit can also exclude. What distinguishes this from a coercive scheme is structure, not intent — replicating the technology without these conditions produces a different system entirely.
Elected by the household, never mandated.
The state does not author the code; an independent body does.
No third party is compelled to suppress speech outside the namespace.
The open internet continues unchanged, in parallel, in full.
Policy authority is independent, with published criteria and appeal.
A household may discontinue at any time, without penalty.
Earlier efforts to set aside a safer part of the namespace — kids.us, .xxx, .kids — were serious attempts by serious people, and the public record of each is worth reading. Between them they leave three questions open.
Authority — on what basis the rules of the space are set. Distribution — how the space is actually reached. Verification — how conformity is checked after entry, and by whom. The argument here is that the three have to be answered together.
The foundation maintains a single canonical metadata schema and ratifies changes to it. National trials propose refinements upward; the foundation resolves them into one version — the discipline that prevents every national effort from becoming an unrelated experiment.
The schema is invariant — fields, structure, semantics, the encoding of an admission decision — in every jurisdiction that adopts it. The values inside those fields are local: what counts as suitable for an age band, what the admission floor requires, stays domestic content judgment.
The analogy is ISO 8601. The date format is identical everywhere on earth; what is scheduled on that date is entirely local. The foundation owns the format — the nation owns the policy applied through it.
The admission code is an objective minimum. Above it, families layer higher standards through third-party curators according to their own values — a faith-based curator, an educational curator, a values-based curator can all operate on the same namespace without fragmenting the compliance burden, and without the schema itself classifying content by any protected characteristic.
Roles the model separates. Nothing here describes a body in operation.
The International Foundation for Child-Safe internet Standards maintains the metadata schema and the model admission code for governed, child-safe namespaces. It does not operate a registry, does not make substantive conformity judgments about individual material, and does not set any jurisdiction's content policy.
That separation is deliberate and structural: the clearest lesson from earlier attempts is that a policy body which depends on a registry operator eventually serves it.
The schema is published openly so conformance can be checked by anyone — including by people who disagree with it. Policy authority sits with the foundation rather than with any registry operator, and the criteria are written down so they can be read, compared and contested.
S. Harrison Knudson, J.D., is Director of the Foundation. His work applies land-use and planning law — the legal technology of admission, placement and conformity — to internet governance, and he is the author of the foundation's working papers on additive namespace architecture.
Correspondence is welcome and is best conducted in writing — see Contact.
No: participation is elected by the household, never mandated; the open internet continues unchanged in parallel; nothing is taken down anywhere; and governance is designed to be independent, with published criteria and a right of appeal. Remove any one condition and it becomes a different, coercive system — which is exactly why all six are written down.
The standard governs what is inside the namespace; it does not claim to prevent someone from leaving it. That's a coverage boundary, not a system failure — a library's collection policy doesn't fail because a child borrows a book from a friend.
Age verification asks "how old is this person?" at every gate on the open web — expensive, privacy-invasive, and it still doesn't tell you what a space's rules are. Admission asks "what are the rules of this space?" once, at entry, and lets a publisher choose to meet them.
Each jurisdiction does, for itself. The foundation standardises the format the decision is expressed in — the same discipline ISO 8601 applies to dates — not what the decision says.
Nothing about the format guarantees justice on its own — but published, machine-readable criteria are visible, comparable across jurisdictions, and quotable by anyone who wants to contest them. An exclusion written down is easier to challenge than one that never was.
Everything here is released for comment. Critical reactions — technical, legal, governance-related or empirical — are more useful to us than supportive ones.
Each entry below has a one-page brief you can read here on the site. The full paper is sent on request, so that it can be matched to the jurisdiction and the question you are working on. Work under submission to journals and other publications is not listed.
One page now; the paper on request. Zoning the Namespace, the theoretical root the rest of the series assumes, is in preparation and available on request in the meantime.
AI systems now occupy relational roles with children — tutor, confidant, first responder to a disclosure. Roles of that kind have always carried professional duties. Declining to answer is not the same as getting a child to someone who can help, and current deployments do not meet the standard the role implies.
A least-cost-avoider analysis of child-safety cost allocation. Who is best placed to prevent the harm, who is actually bearing the cost of preventing it, and what follows when those are not the same party. The answer is not that anyone is acting in bad faith; it is that the burden has settled in the wrong place.
An identification problem. Demand for child-safety provision has been inferred from instruments that move three variables at the same time, and the confound has never been isolated. The paper argues for what would have to be measured to know, and is careful not to convert that argument into a prediction.
Shorter work written for a particular audience or question.
A school has a registrar and an address. Enrollment asks who may enter; siting asks what a child crosses to get there. Age assurance is the registrar and says nothing about the environment. Where content was never admitted, there is nothing inside to circumvent — a property of construction rather than a claim about detection accuracy.
At a liquor store the customer proves age: one bit, tested once. At a school the adult proves fitness — checked, credentialed, trained, renewed, revocable — and the child proves nothing. Children's spaces need the second regime. What that implies for the systems now being placed in front of children.
Written for registry operators. What a protected area inside a national domain would actually require to run — admission decisions, appeals, eviction, error rates — and why the operator is the only party positioned at the right layer to hold it.
Introductory material, free to download, copy and circulate.
Age assurance asks who is entering. Admission asks what is allowed inside. Why an interoperable child-safety metadata standard should be designed in Africa rather than built elsewhere and brought there.
The short introduction, read against the regimes now in force: the Online Safety Act, the Age Appropriate Design Code and the Digital Services Act. Sets out why the household election rests where Article 5 of the Convention on the Rights of the Child already places it, and the five conditions that separate this from privatised censorship.
A short guide for advocates. What to ask your government for, in one sentence, and why a protected area inside your own national domain needs no global permission to create. Written in plain terms, to be handed to someone who has never heard the argument.
The longer documents carry material that is specific to a jurisdiction, a sector or an institution — and in some cases organisational proposals that only make sense once we know who is reading. Rather than post a version written for nobody in particular, we send the one that fits. Tell us the audience and the question in the request and we will reply with the relevant document, usually within a few days. Requests are welcome from critics as readily as from allies.
The foundation maintains further briefs and material written for particular jurisdictions, sectors and audiences, alongside work under submission to journals and other publications and so not published on this page. Any of it is available on request. Write to shknudson@ifcsis.org describing the audience and the question you are working on, and we will send what fits.
The controlled vocabulary and machine-readable schema are maintained separately at standard.ifcsis.org. Comments and corrections: shknudson@ifcsis.org
Written enquiries are welcome. Please include the jurisdiction your enquiry concerns.