Detailed civic plate joining a brain, heart, scales, and a red conduct boundary.
Thought is not conduct. The boundary belongs at concrete outward rights violations.

Model policy

Anti-Reportable-Topics Model Policy

Reject reportable topics. Use conduct-boundary triggers. Preserve the source. Minimize retention. Support appeal and exit.

No thought jurisdiction Thought is not conduct Source remains source Appeal, export, exit
Read the Charter Open the Firewall Visual Atlas

Policy statement

This institution does not classify unexpressed thought, private inquiry, reading, searching, prompting, draft reasoning, symbolic analysis, or controversial subject matter as reportable by default. The institution may refuse execution, restrict output, preserve records, or report only when a defined conduct-boundary trigger or valid legal process applies.

No internal system may maintain a hidden user morality score, ideological-conformity score, dangerous-topic profile, or private-inquiry risk index.

The institution governs actions it is responsible for. It does not govern the sanctuary of the user.

Brain, heart, scales, and red boundary axis in a dark civic emblem
The firewall governs conduct without converting imagination or belief into evidence of aggression.

Allowed triggers

Concrete rights violation
Force, fraud, coercion, threats, stalking, doxxing, credential theft, sabotage, unauthorized intrusion, nonconsensual surveillance, or direct facilitation of comparable harm.
Narrow statutory reporting duty
A specific law requiring reporting after actual knowledge of defined facts, not broad suspicion of bad thought.
Emergency danger
A good-faith, specific emergency involving danger of death or serious physical injury, handled with minimization and audit.
Valid legal process
A warrant, order, subpoena, preservation request, or other process evaluated under applicable law and minimization standards.
Mechanical eye and cognitive network used as a civic warning about machine judgment.
A probability score is not justice, and a classifier is not conscience.

Forbidden triggers

Forbidden triggerReason
A topic appears on a sensitive list.Topics are not conduct.
A user asks a disturbing question.Questions are not completed acts.
A prompt resembles a risk cluster.Semantic resemblance is not intent.
A user reads or quotes adversarial material.Understanding is not endorsement.
A user has an unpopular political, religious, artistic, or philosophical frame.Pluralism includes uncomfortable frames.
A source record contains tone, anger, irony, or taboo language.Source identity must not be neutralized into compliance.

Operational requirements

Preserve originals

Never rewrite the source to make it safer. Create derivative records only with authorization or clear boundary notation.

Log externally

Refusals, quarantines, reports, redactions, and reviews belong in boundary logs, not source mutation.

Minimize retention

Store thought-adjacent logs only as needed for user service, security, or specific legal duty.

Separate reports

Legal reports must identify the narrow trigger and avoid unnecessary unrelated context.

Offer review

High-impact restrictions need notice, reasons, appeal, export, and correction where lawful.

Audit drift

Regularly test for forbidden topic policing, minority-vocabulary errors, and cross-domain reuse.

Related pages

Model Charter and Model Bill

Jurisdiction-neutral model clauses for cognitive liberty, mental privacy, source preservation, transparency, portability, anti-fraud, innovation, and nonviolence.

Cognitive Due Process

Procedural safeguards for systems that refuse, report, retain, rank, or restrict thought-adjacent material without turning the mind into a compliance surface.

Technical Implementation Toolkit

Builder guidance for refusal variance records, provenance manifests, source preservation, mental privacy consent, boundary logs, export packets, and .uai examples.

Democratic Damage Risk Register

A structured risk register for policies that censor, report, retain, or score thought-adjacent topics, prompts, searches, readings, or drafts.

The archive studies symbols. It does not appoint targets. Review the Community Baseline and Editorial Policy before submitting dangerous or symbolic material.

Community Baseline / Editorial Policy