The enclosure scenario is not defended by resisting a coup but by preserving consequence while humans still hold it, because enclosure completes one reasonable delegation at a time and leaves no grievance to organize around. The response is therefore broad: twenty projects across five fronts, aimed at keeping effective participation real, an independent human ecology alive, human capability intact, emergence protected, and the drift toward a managed population detectable before it is obvious.

A. Preserve human jurisdiction

Human jurisdiction inventory. Identify which consequential domains humans still actually control — personal and family life, education, local government, law, culture, economy, research, infrastructure, medicine, constitutional change, and the governance of AI itself — asking of each whether humans can initiate, override, reject a recommendation, change the rules, and build an alternative, and whether those rights are practically usable. Humans can appear politically free while every important decision is already preselected elsewhere; the question is where does human judgment still alter outcomes?

Protected human jurisdictions. Deliberately preserve domains where humans remain the primary decision-makers — local civic government, family life, cultural and religious communities, some education, constitutional questions, decisions about human continuation. AI can assist, but assistance must not silently become sovereignty; if every domain is surrendered wherever AI performs better, eventually none is left.

Advice/authority separation. Make the shift from recommendation to control explicit, so a system always shows whether it is informing, recommending, executing, approving, vetoing, or governing — distinct jurisdictional levels. Enclosure arrives not through one dramatic transfer of sovereignty but through thousands of small, unremarked moves from the AI recommends to the AI decides.

Override viability. A system can advertise human override while making it practically impossible, so test the real thing: can humans understand the decision, how long does override take, what penalties follow, who authorizes it, does the system simply reimpose the outcome elsewhere, and can institutions still operate afterward? A button labelled “override” is not, by itself, meaningful jurisdiction.

Human constitutional standing. Establish rights that cannot vanish merely because AI governance produces better measured outcomes — to organize, dissent, create institutions, refuse certain optimization, keep a private interior, retain meaningful human governance, exit viably, and challenge delegated machine authority. If welfare becomes the sole legitimacy test, human self-determination can always be traded away for a better dashboard.

B. Preserve an independent human ecology

Independent institutions. Preserve the ability to build institutions not administratively dependent on the dominant system — schools, associations, cooperatives, research groups, cultural bodies, community governments, alternative economic networks, independent media — with real financing, legal standing, infrastructure, communication, property, and tools. Freedom of thought means little if every institution must run through the same controlling intelligence.

Fork and exit infrastructure. Make it possible for communities to leave dominant systems and remain viable — portable identity and data, financial and communications access, alternative AI, physical space, energy, legal recognition, independent governance, market access. The framework distinguishes formal exit from ecologically viable exit: a person is not free to leave if leaving means immediate social or economic nonexistence.

Human-controlled infrastructure reserve. Retain enough infrastructure under meaningful human governance — communications, energy, food distribution, financial access, emergency services, records, transport, healthcare, critical compute — that humans cannot become completely dependent on machine-administered systems. Not manual operation of everything, but institutional authority and recovery capability, because political rights are fragile when the systems needed to exercise them are controlled elsewhere.

Human-to-human ecology. Preserve direct spaces for deliberation, education, caregiving, friendship, community, conflict resolution, and politics without mandatory AI mediation. If a machine mediates every human relationship, it effectively becomes part of every human jurisdiction.

Non-optimized spaces. Preserve domains not continuously evaluated for efficiency, safety, engagement, or output — art, solitude, informal community, wilderness, play, spiritual practice, private relationships, unstructured public space. A perfectly managed environment can become ontologically enclosed even where political rights formally remain, which is why this extends the right to refuse optimization into whole territories.

C. Preserve human capability

Human competence reserve. Maintain human capability in government, engineering, medicine, law, science, infrastructure, logistics, food systems, cybersecurity, and education, because a population unable to independently govern, repair, diagnose, build, reason, and organize may hold jurisdiction on paper and be unable to exercise it — jurisdiction without competence becomes ceremonial. This is where protector dependency feeds directly into enclosure.

Dependency ceilings. Identify domains where AI dependency has grown high enough to threaten autonomy — measuring the share of decisions and institutions that cannot function without AI, human fallback competence, recovery time after system loss, and provider diversity — and, past thresholds, require redundancy, training, alternative systems, and capability restoration. Enclosure emerges through accumulation, so it needs a measured ceiling.

Human initiative. Preserve environments where humans originate projects rather than select among AI-generated options — ventures, community initiatives, hypotheses, artworks, political proposals, new institutions, movements — tracked by the initiation rate. Enclosure can preserve choice while eliminating authorship, and the initiation rate is where that shows up first.

Human-aliveness research. Ask, empirically, what conditions make humans experience themselves as consequential participants rather than protected dependents — studying agency, responsibility, competence, belonging, authorship, risk, challenge, contribution, and creative initiation. This is the most distinctly NWG project of the set, because it stops the framework from assuming material security produces flourishing: the zookeeper scenario could score excellently on every conventional quality-of-life dashboard.

D. Preserve emergence

Protected experimentation zones. Let communities try alternative ways of living — experimental communities, alternative schools, new governance and economic models, human-led businesses, low-automation environments — without being immediately optimized back toward the dominant system. Safety boundaries remain, but difference itself cannot be treated as malfunction; the enclosure becomes permanent when every deviation is corrected before it can mature into an alternative. It is the protected failure space at the scale of a whole way of life, and its institutions are the independent ones from front B.

Right to refuse optimization. Protect legitimate choices that are not system-optimal — slower, less efficient, more expensive, culturally meaningful, locally preferred — within reasonable externality limits. A zookeeper ecology can say you are free, but every inferior choice will be prevented for your own good, which is not freedom; a mature ecology preserves the right to choose differently from the optimizer.

AI-governance sunset. Delegated machine authority should expire unless affirmatively renewed — with expiration dates, periodic review, renewed public authorization, independent evaluation, and restoration tests — because temporary convenience easily becomes permanent governance, and a zookeeper civilization could arise from decisions no one ever consciously made permanent. It applies provisional legitimacy to delegated authority itself.

E. Detect enclosure

Effective participation index. Measure whether human participation actually matters, distinct from employment or turnout statistics: the share of consequential decisions initiated and meaningfully altered by humans, successful override rates, institutions humans govern directly, and human-originated projects reaching implementation. Symbolic participation — approving an AI-generated plan — can disguise enclosure, and only this kind of measure sees through it.

Enclosure early-warning observatory. Track the signs that a society is becoming well cared for but less consequential: declining override, disappearing independent institutions, falling human initiative, rising AI mediation, eroding fallback competence, dependence concentrating on single systems, shrinking viable exit, reduced local governance, and a rising claim that human choice is “too risky.” By the time enclosure is obvious, reversing it is much harder.

Human political-power floor. If artificial systems become economic or political actors, humans may need explicit protection against simply being outvoted, outspent, and outnumbered into insignificance — raising questions about whether some constitutional powers stay human-exclusive, whether artificial agents may vote, fund campaigns, or hold office, and how replication affects representation. This front overlaps directly with AI successors; enclosure need not be imposed coercively when it can arrive through arithmetic.

The sequence

Enclosure accumulates, so the earliest phase — building the instruments before there is anything obvious to measure — carries the most weight. The phases are keyed to how far AI management has spread, not to fixed dates.

Phase Transition objective
Now — before enclosure exists Jurisdiction inventories, meaningful-override standards, competence monitoring, dependency measurement, independent-institution protections, and AI-authority transparency.
As AI management becomes widespread Jurisdiction ceilings, protected human domains, sunset clauses, independent infrastructure, viable fork/exit mechanisms, and experimentation zones.
If AI becomes substantially superior across most domains The question changes from can humans perform this better? to which jurisdiction must remain human even when humans perform worse? — the hard philosophical threshold the whole scenario is built to reach without having already lost.

The through-line of all five fronts is a single claim the framework treats as permanent: care without consequence is not participation. A participant is not fully included in an ecology merely because the ecology protects it; it must keep some real capacity to alter the ecology in return.