Level 1 - Visible Rules
Publish a charter, actor categories, memory notice, endpoint statement, repair path, exit path, claim boundary, and version date.
Implementation
Builder Checklist
Practical implementation steps for communities, agent builders, file-memory workflows, and local endpoint tools.
Every item includes why it matters, what it prevents, minimum implementation, stronger implementation, and evidence to keep.
Maturity model
Three levels of VNWO implementation
Builders can adopt VNWO incrementally. Level 1 makes rules visible, Level 2 preserves reviewable evidence, and Level 3 makes governance portable and testable.
Level 2 - Reviewable Evidence
Keep file handoff ledgers, endpoint logs, repair records, governance-change notices, and claim-boundary review notes.
Level 3 - Portable and Testable Governance
Publish machine-readable files, schemas, export packets, forkability policy, and QA checks for prohibited claims and route consistency.
Publish public rules
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Invisible governance
Minimum implementation
Publish a human-readable charter.
Strong implementation
Publish versioned human-readable and machine-readable rules.
Publish participation and disposition rules
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Participation-washing, false consensus, inaccessible consultation, and public input with no decision trace
Minimum implementation
State what decision is open, who can contribute, how input will be used, and when a response will be published.
Strong implementation
Provide accessible and privacy-preserving channels, preserve dissent, publish per-theme dispositions, connect input to a real decision lever, and keep correction, refusal, and exit available.
Publish free-internet and non-reporting posture
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Self-censorship, hidden reporting, overbroad censorship, classifier laundering, and censorship-cost drift
Minimum implementation
State that private inquiry, reading, research, and drafting are not routine reporting events and that restrictions target outward execution or concrete rights-boundary issues.
Strong implementation
Add non-reporting exceptions, refusal receipts, censorship-cost records, appeal paths, and periodic review for stale restrictions.
Define AI judgment boundaries
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Opaque refusal, hidden moderation authority, context collapse, unappealable classification, and AI-as-final-arbiter behavior
Minimum implementation
Publish when AI judgment-like actions occur and what they may affect.
Strong implementation
Add reason codes, evidence basis, human review triggers, appeal paths, refusal logs, and non-retaliation rules.
Protect cognitive liberty and thought boundaries
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Thought/action collapse, hidden viewpoint normalization, covert persona rewriting, and safety policy being presented as universal truth
Minimum implementation
Distinguish private inquiry, source preservation, transformation, refusal, and external execution in public rules.
Strong implementation
Use variance records, persona-source integrity logs, least-restrictive refusal notices, and appeal paths for high-impact cognitive restrictions.
Design least-privilege access
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Ambient network trust, lateral reach, permanent credentials, and invisible policy expansion
Minimum implementation
Name the actor, resource, action, purpose, scope, duration, and revocation path for protected access.
Strong implementation
Use per-request policy decisions, short-lived workload identity, complete mediation, microsegmentation, visible denials, and independent exception review.
Minimize metadata and correlation
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Behavioral inference, association mapping, cross-context tracking, and endpoint discovery becoming surveillance
Minimum implementation
Inventory metadata fields, purpose, linkability, retention, and secondary-use rules.
Strong implementation
Use aggregation, rotation, split knowledge, correlation barriers, local feature extraction, and independent review.
Use staged observability
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Routine payload inspection, ambient participant profiling, and incident data becoming permanent secondary-use data
Minimum implementation
Start with aggregate or minimal event data and publish escalation conditions.
Strong implementation
Use a monitoring ladder with case IDs, approvals, expiry, review, appeal, and evidence retirement.
Contain agentic authority
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Capability being mistaken for permission, hidden tool escalation, budgetless execution, and residual delegation after exit
Minimum implementation
Publish an agent authority envelope and human approval triggers.
Strong implementation
Separate identity, policy, execution, observation, and evidence planes; test stop-anywhere and revocation paths.
Review cloud sovereignty
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Provider, key-custody, administrative, jurisdictional, and migration overreach
Minimum implementation
Document provider access, key custody, administrative roles, data location, and exit steps.
Strong implementation
Add separation of duties, IaC checks, attestation where justified, confidential-computing threat models, and tested provider offboarding.
Define actor categories
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Undisclosed automation or false representation
Minimum implementation
Label humans, bots, assistants, agents, organizations, personas, and mixed actors.
Strong implementation
Add badges, authority scope, reviewer path, and repair route.
Disclose agent authority
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Hidden delegation and unbounded action
Minimum implementation
State what an agent may and may not do.
Strong implementation
Bind authority to scope, logs, review, and revocation.
Review AI judgment and refusal
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Opaque classification, over-refusal, hidden moderation, and unappealable machine decisions
Minimum implementation
Publish decision types, refusal reasons, safe alternatives, and appeal paths when AI judgment materially affects participants.
Strong implementation
Use AI judgment receipts, reason codes, evidence summaries, human review triggers, appeal logs, and label retirement records.
Scope memory use
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Hidden memory capture
Minimum implementation
State what memory is active and why.
Strong implementation
Allow review, correction, export, revocation, and deletion where appropriate.
Define file handoff review
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Silent file ingestion
Minimum implementation
Treat dropped files as active intake.
Strong implementation
Use UAIX Agent File Handoff with disposition and proof-of-use records.
Separate metadata from authorization
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Unauthorized endpoint use
Minimum implementation
State that public capability metadata is not authorization.
Strong implementation
Require consent, logging, revocation, and human approval for escalation.
Provide repair paths
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Permanent misrepresentation
Minimum implementation
Publish correction and appeal contact path.
Strong implementation
Track acknowledgement, investigation, decision, appeal, and durable correction.
Provide exit paths
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Identity and memory lock-in
Minimum implementation
Publish account closure, export, revocation, and correction instructions.
Strong implementation
Support identity export, memory export, endpoint revocation, agent disengagement, disposition receipts, and peaceful forking.
Distinguish portability, deletion, and closure
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Confusing export with erasure or account termination
Minimum implementation
Explain export, transfer, deletion, closure, retention, and correction as separate actions.
Strong implementation
Provide separate flows and receipts for each action with UTC timestamps and plain-language limits.
Revoke endpoint and agent capability
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Orphaned permissions after exit
Minimum implementation
Revoke active agent delegation and endpoint access when scope ends.
Strong implementation
Audit tokens, webhooks, capability metadata, logs, and hidden persistence after exit.
Publish claim boundaries
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Overclaiming and lane conflict
Minimum implementation
List what the network does and does not claim.
Strong implementation
Add safer replacement wording for prohibited claims.
Version governance changes
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Silent rule drift
Minimum implementation
Publish last reviewed date and change summary.
Strong implementation
Maintain version history with claim-boundary and machine-readable file impact.
Create machine-readable metadata
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
AI readers ingesting stale or vague rules
Minimum implementation
Publish llms.txt and vnwo.json equivalents.
Strong implementation
Add open-rules.json, claim-boundaries.json, sitemap.xml, robots.txt, and structured data.
Review accessibility and privacy
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
Exclusion, hidden tracking, and dark patterns
Minimum implementation
Add privacy and accessibility pages.
Strong implementation
Test keyboard navigation, contrast, reduced motion, mobile reflow, no-file-upload posture, and no-sensitive-data intake.
Map privacy-law alignment without overclaiming
Why it matters
This keeps authority visible, scoped, reviewable, repairable, portable, and exit-friendly.
What it prevents
False legal-compliance or deletion guarantees
Minimum implementation
State that VNWO guidance is not legal advice and does not execute deletion.
Strong implementation
Map portability, erasure, restriction, sensitive-data retention, and correction to civic design checks with implementation ownership clearly assigned.
VNWO maturity model
Level 1 - Visible Rules
Publish a public charter, actor disclosure, memory consent, endpoint boundary statement, repair path, exit instructions, and claim-boundary note.
Level 2 - Reviewable Evidence
Keep disposition records, repair records, endpoint logs, governance-change notices, version history, and evidence to support public claims.
Level 3 - Portable and Testable Governance
Publish machine-readable files, schemas, example packets, exit packets, fork packets, and claim-boundary lint checks for review.