Complete implementation kit · KIT-PUBLIC-INSTITUTIONS-001

Public authority must remain answerable to the people whose minds it can affect.

Public institutions protect cognitive liberty when they require individualized or properly bounded authority, minimize data, separate identity from inquiry, disclose government-platform interactions, constrain automated decisions, preserve anonymous and pseudonymous access, provide human review and remedy, delete unsupported inferences, and terminate emergency powers when the defined need ends.

Direct implementation answer

Public institutions protect cognitive liberty when they require individualized or properly bounded authority, minimize data, separate identity from inquiry, disclose government-platform interactions, constrain automated decisions, preserve anonymous and pseudonymous access, provide human review and remedy, delete unsupported inferences, and terminate emergency powers when the defined need ends.

People and roles

Do not collapse different people into one surveillance category.

The same technology can create different risks depending on power, age, role, location, legal authority, and the consequences of disclosure.

KIT-PUB-POP-01

Residents and service users

People seeking benefits, licenses, education, health information, public records, library access, or ordinary services should not have unrelated lawful inquiry converted into an eligibility or risk profile.

KIT-PUB-POP-02

Children and adolescents

Safeguarding must address exploitation and concrete harm while preserving age-appropriate confidential help-seeking, reading, identity development, and voluntary inquiry.

KIT-PUB-POP-03

People without standard identity documents

Essential information and services need accessible alternatives for refugees, migrants, unhoused people, survivors, and others who cannot safely or readily present a conventional credential.

KIT-PUB-POP-04

Public servants and contractors

Employment security and operational integrity do not justify continuous ideological, emotional, religious, health, or cognitive monitoring of workers.

KIT-PUB-POP-05

Journalists, researchers, advocates, and whistleblowers

Contact with controversial people, use of encryption, sensitive searches, public-records requests, and investigation of alleged wrongdoing are protected functions, not automatic indicators of threat.

KIT-PUB-POP-06

Targets, witnesses, and uninvolved associates

Investigative authority must distinguish the person or event supported by evidence from family, colleagues, readers, visitors, bystanders, or communities swept into the same data graph.

KIT-PUB-POP-07

Political, religious, and cultural minorities

A system must not convert minority belief, dissent, satire, protest, association, language, or historical interpretation into a durable adverse classification.

KIT-PUB-POP-08

Judges, auditors, ombuds, and elected overseers

Independent reviewers need complete evidence and secure access while remaining separate from operational chains that benefit from expanding the power under review.

Context boundaries

Controls must stay inside the context that justifies them.

KIT-PUB-CTX-01

Routine public administration

Ordinary service delivery should use the minimum data necessary for the transaction and should not silently inherit law-enforcement or intelligence purposes.

KIT-PUB-CTX-02

Individualized investigation

A properly authorized investigation tied to evidence of a defined event is different from reverse-keyword, geofence, bulk, or population-wide collection that begins with no particularized subject.

KIT-PUB-CTX-03

Emergency response

Immediate action may be justified to protect an identifiable person from concrete harm, but emergency access must be logged, reviewed, narrowed, and terminated.

KIT-PUB-CTX-04

Automated eligibility and risk decisions

A tool that helps route a case is different from a model that determines benefits, licensing, employment, education, border treatment, policing, or security consequences without contestable human judgment.

KIT-PUB-CTX-05

Government communication with platforms

Public persuasion, lawful notice, binding order, informal request, procurement leverage, and regulatory threat are different forms of power and must not be blended into an opaque moderation channel.

KIT-PUB-CTX-06

Public records and historical archives

Transparency, privacy, legal holds, classified material, correction, deletion, provenance, and long-term memory require different rules; one cannot be invoked to erase the others.

KIT-PUB-CTX-07

Public procurement and delegated power

A vendor does not reduce government responsibility. Contracted identity, AI, analytics, moderation, or surveillance systems remain public exercises of authority when they affect rights or services.

Bounded threat model

Name systems, protected activity, and credible failure paths.

Scope: Government acquisition and use of search, browsing, AI-prompt, library, communications, location, biometric, neural, and behavioral-inference data; warrants and administrative demands; reverse searches; public records and archives; government-platform communication; automated eligibility and enforcement; public-sector AI procurement and alignment; identity and age verification; emergency powers; national security; and institutional oversight.

Protected activity: Lawful thought, reading, searching, AI inquiry, religion, political belief, association, petition, journalism, legal and historical research, academic work, health inquiry, controversial rights research, anonymous or pseudonymous speech, encryption, privacy tools, and criticism of public authority.

KIT-PUB-TH-01Worldwide principle

Bulk access to cognitive and behavioral data

Search, prompt, library, communications, location, biometric, neural, and inferred-trait records can be acquired or combined without individualized evidence, creating a map of lawful intellectual life.

KIT-PUB-TH-02Jurisdiction-specific legal question

Reverse-keyword and reverse-location investigation

A government may begin with a search term, location, website, or topic and ask a provider to identify everyone who matched it, reversing particularized suspicion.

KIT-PUB-TH-03Worldwide principle

Opaque government pressure on information intermediaries

Informal requests, regulatory threats, procurement leverage, grants, licensing, and public accusations can induce platforms to suppress lawful material without a public order or appeal.

KIT-PUB-TH-04Worldwide principle

Automated eligibility and adverse action

Models used for benefits, licensing, employment, education, border, policing, fraud, or security can convert weak correlations and lawful inquiry into material consequences.

KIT-PUB-TH-05Technical recommendation

Universal identity and attributable access

Age, eligibility, anti-fraud, or service requirements can become a common identifier linking every question, visit, payment, association, and public interaction.

KIT-PUB-TH-06CognitiveLiberties.com policy proposal

Emergency and national-security ratchets

Temporary powers created for war, terrorism, public health, civil disorder, elections, misinformation, or child protection can persist, broaden, and migrate into routine administration.

KIT-PUB-TH-07CognitiveLiberties.com policy proposal

Government alignment of AI and official machine truth

Procurement, licensing, training-data rules, content filters, evaluation criteria, or informal pressure can force AI systems to reproduce state-preferred narratives without visible attribution.

KIT-PUB-TH-08Operational practice

Public-record manipulation and institutional amnesia

Deletion, classification, altered metadata, broken provenance, inaccessible archives, and vendor withdrawal can prevent later accountability or rewrite the public record.

KIT-PUB-TH-09Worldwide principle

Secondary use and downstream propagation

Data collected for one service can be reused for enforcement, intelligence, employment, fraud, model training, or political profiling and copied into systems the person cannot see.

KIT-PUB-TH-10Operational practice

Oversight capture and unverifiable secrecy

Review bodies may lack technical access, independence, budget, standing, adversarial evidence, or authority to stop a system whose operators define their own compliance.

Minimum safeguards

Build a rights-preserving floor before adding complexity.

These safeguards state the purpose that must survive local implementation. They are not a claim that every jurisdiction uses identical law or procedure.

KIT-PUB-SG-01Worldwide principle

Presume lawful inquiry is not evidence of wrongdoing

Do not treat a question, topic, reading choice, source contact, protest interest, encryption use, or controversial rights research as proof of belief, intent, dangerousness, or unfitness without independent evidence.

KIT-PUB-SG-02Worldwide principle

Require defined harm and bounded authority

Every restrictive action must name the concrete harm, legal authority, evidence threshold, affected population, scope, duration, responsible official, reviewer, remedy, and end condition.

KIT-PUB-SG-03Jurisdiction-specific legal question

Use individualized or properly bounded process

Prefer particularized warrants, orders, and decisions. Where another legal standard applies, document why it is lawful, necessary, proportionate, and no broader than the actual investigation or service.

KIT-PUB-SG-04Technical recommendation

Minimize and separate identity from inquiry

Collect only the identity required for the transaction and keep credentials separate from search, reading, prompt, library, and content records through scoped identifiers and split trust.

KIT-PUB-SG-05CognitiveLiberties.com policy proposal

Offer anonymous and pseudonymous public access

Preserve non-account reading, public information, public records, library use, complaint intake, and other lawful inquiry where full civil identity is not necessary.

KIT-PUB-SG-06Technical recommendation

Separate request-level safety from person-level judgment

A warning, refusal, rate limit, or local safeguard should ordinarily attach to the request. Persistent risk labels, referrals, or eligibility consequences require a higher and contestable evidentiary process.

KIT-PUB-SG-07Operational practice

Make government-platform interactions reviewable

Record the sender, recipient, legal basis, requested action, urgency, content category, response, appeal path, retention, and whether the communication was persuasion, notice, request, or binding order.

KIT-PUB-SG-08CognitiveLiberties.com policy proposal

Require human review before material impairment

Benefits, licensing, discipline, employment, border, policing, fraud, and security decisions must not become final solely because a model generated a score or label.

KIT-PUB-SG-09Worldwide principle

Provide notice, evidence, appeal, and correction

Explain the rule and material facts, identify the machine role, permit contextual evidence, assign an independent reviewer, set deadlines, pause avoidable harm, and correct downstream systems.

KIT-PUB-SG-10CognitiveLiberties.com policy proposal

Delete unsupported cognitive data and inferences

Remove raw records, embeddings, labels, scores, watch-list entries, model features, training copies, and propagated derivatives when authority or need ends or the judgment is overturned.

KIT-PUB-SG-11Technical recommendation

Use privacy-preserving safety architecture

Prefer local processing, oblivious transport, private retrieval, anonymous credentials, threshold authorization, private aggregation, and synthetic testing before centralized identity-linked monitoring.

KIT-PUB-SG-12Operational practice

Protect public memory with provenance and plural custody

Maintain fixity, provenance, correction histories, lawful access, format migration, independent copies, and clear rules for privacy, classification, legal holds, and eventual release.

KIT-PUB-SG-13CognitiveLiberties.com policy proposal

Build automatic anti-ratchet controls

Use scope-limited appropriations, technical purpose separation, expiration, reauthorization with evidence, public statistics, independent audit, deletion, and tested dismantling procedures.

KIT-PUB-SG-14Worldwide principle

Preserve independent oversight and whistleblowing

Protect inspectors, auditors, courts, ombuds, legislators, journalists, researchers, and whistleblowers from retaliation and give them the evidence needed to test official claims.

Staged implementation

Move from visibility to enforceable controls to durable resilience.

First 30 daysKIT-PUB-STAGE-30

Map authority, data, and immediate rights exposure

Outcome: The institution can identify every cognitive-data flow, restrictive power, automated decision, platform communication, emergency authority, vendor, owner, retention rule, and route for urgent human review.

  1. KIT-PUB-30-01Operational practice

    Name accountable owners

    Assign executive, legal, privacy, security, records, procurement, accessibility, civil-rights, and independent-review owners with published responsibilities.

    EvidenceCL-14-02
  2. KIT-PUB-30-02Operational practice

    Inventory cognitive and behavioral data

    Map search, prompt, browsing, library, communications, location, biometric, neural, identity, inferred-trait, risk, and eligibility data from collection through deletion.

  3. KIT-PUB-30-03Jurisdiction-specific legal question

    Inventory coercive and restrictive authorities

    List warrants, subpoenas, administrative demands, emergency access, reverse searches, secrecy orders, content requests, licensing, procurement, grants, and platform contacts.

  4. KIT-PUB-30-04CognitiveLiberties.com policy proposal

    Freeze unsupported high-impact automation

    Pause new person-level inference, automated adverse action, emotion or ideology scoring, and bulk query analysis until authority, validity, necessity, and remedy are documented.

  5. KIT-PUB-30-05Operational practice

    Publish immediate rights and escalation boundaries

    Tell the public what is collected, what is not, how emergency access works, who can challenge a decision, and how to reach a human without disclosing more sensitive information.

Within 90 daysKIT-PUB-STAGE-90

Convert principles into enforceable controls and remedies

Outcome: The highest-risk powers have written thresholds, technical minimization, contract controls, human review, transparency, appeal, deletion, oversight, and tested incident procedures.

  1. KIT-PUB-90-01CognitiveLiberties.com policy proposal

    Adopt a cognitive-liberty impact assessment

    Require a documented assessment before identity gates, monitoring, model alignment, automated decisions, reverse searches, content restrictions, or new cognitive-data uses.

  2. KIT-PUB-90-02Technical recommendation

    Implement minimization and purpose separation

    Change schemas, access roles, logs, tokens, and storage so service delivery, analytics, enforcement, intelligence, training, and public records do not silently share one data lake.

  3. KIT-PUB-90-03Operational practice

    Standardize government-platform records

    Create machine-readable request and order records, public aggregate reporting, legal review, conflict escalation, and retention rules.

  4. KIT-PUB-90-04Worldwide principle

    Create independent appeal and correction

    Provide timely human reconsideration, access to material evidence, representation or support where needed, downstream correction, restoration, and deletion of unsupported labels.

  5. KIT-PUB-90-05Operational practice

    Exercise emergency and breach procedures

    Run tabletop tests for immediate threats, emergency data demands, mass false positives, vendor failure, archive tampering, model drift, data breach, and unlawful pressure.

Within 365 daysKIT-PUB-STAGE-365

Institutionalize plural oversight, resilience, and automatic limits

Outcome: Cognitive-liberty safeguards survive leadership change through law, budget, architecture, records, procurement, independent oversight, public evidence, and tested exit.

  1. KIT-PUB-365-01CognitiveLiberties.com policy proposal

    Codify rights and prohibited uses

    Establish enforceable limits on query profiling, covert cognitive manipulation, unsupported sensitive inference, secret person-level risk labels, and adverse decisions based solely on lawful inquiry.

  2. KIT-PUB-365-02Operational practice

    Create durable oversight capacity

    Fund independent technical auditors, inspectors, ombuds, courts, legislatures, civil society, and public-interest research with secure evidence access and authority to order correction.

  3. KIT-PUB-365-03Technical recommendation

    Build vendor and infrastructure exit

    Maintain export, interoperability, alternate providers, open formats, domain and key custody, archive copies, migration runbooks, and termination rights.

  4. KIT-PUB-365-04CognitiveLiberties.com policy proposal

    Make emergency authority expire by design

    Tie extraordinary access, filtering, monitoring, and alignment powers to narrow triggers, automatic technical expiration, public reauthorization evidence, deletion, and budget termination.

  5. KIT-PUB-365-05Operational practice

    Publish aggregate accountability evidence

    Report request categories, error and appeal outcomes, deletion performance, audit findings, model changes, emergency uses, and remedy without exposing lawful inquiries or protected people.

Evidence and procurement checklist

Do not buy a promise. Require inspectable evidence.

A policy statement is not proof of system behavior. Require architecture, configuration, tests, contracts, logs with bounded retention, deletion evidence, and a remedy when the provider is wrong.

KIT-PUB-PROC-01Jurisdiction-specific legal question

Purpose and legal authority

Require the exact public purpose, legal basis, affected decision, prohibited uses, and reasons less intrusive alternatives are insufficient.

EvidenceCL-03-02
KIT-PUB-PROC-02Operational practice

Collected fields and derived inferences

Demand every raw field, identifier, metadata element, embedding, score, label, relationship, prediction, and inferred trait.

EvidenceCL-15-01
KIT-PUB-PROC-03Technical recommendation

Identity and attribute architecture

Document whether full identity is necessary, which selective-disclosure or anonymous alternatives were tested, and how identifiers remain unlinkable across services.

KIT-PUB-PROC-04Operational practice

Retention and verified deletion

Specify event-level schedules, legal holds, backup expiry, model and vector deletion, downstream copies, proof of completion, and consequences for failure.

KIT-PUB-PROC-05CognitiveLiberties.com policy proposal

Training, evaluation, and product improvement

Prohibit training or evaluation use unless explicitly authorized, necessary, documented, revocable, and separable from service delivery.

EvidenceCL-32-04
KIT-PUB-PROC-06CognitiveLiberties.com policy proposal

Advertising, sale, profiling, and unrelated reuse

Contractually prohibit commercial targeting, data brokerage, political profiling, sensitive-trait inference, and cross-context use of public-service data.

KIT-PUB-PROC-07Operational practice

Subprocessors and cross-border processing

Name every processor, location, access role, onward transfer, government-demand exposure, and replacement-notice process.

KIT-PUB-PROC-08Operational practice

Model, policy, and ruleset changes

Require advance notice, regression evidence, version retention, rollback, public explanation, and renewed approval for material changes.

EvidenceCL-15-02
KIT-PUB-PROC-09Technical recommendation

Accuracy, validity, and differential impact

Demand evidence relevant to the actual population and decision, uncertainty intervals, false-positive and false-negative costs, language and disability testing, and limits on inference.

KIT-PUB-PROC-10Worldwide principle

Human review and contestability

Require a named human decision-maker, access to material evidence, contextual submission, appeal independence, deadlines, restoration, and downstream correction.

KIT-PUB-PROC-11Jurisdiction-specific legal question

Government-demand handling

Require authentication, jurisdiction review, scope narrowing, emergency validation, secrecy-order review, challenge rights, notice rules, and disclosure statistics.

KIT-PUB-PROC-12Technical recommendation

Security, encryption, and administrator authority

Document encryption, key custody, privileged roles, recovery, logging, insider controls, breach testing, and whether the provider can read protected content.

KIT-PUB-PROC-13Operational practice

Accessibility and non-digital alternatives

Require equivalent access for disability, language, connectivity, documentation, and device constraints without converting accommodation into a risk signal.

EvidenceCL-42-04
KIT-PUB-PROC-14Jurisdiction-specific legal question

Public records, audit logs, and archival custody

Define which records document official action, which data remain private, how corrections are preserved, when classification ends, and how provenance and fixity are tested.

KIT-PUB-PROC-15Operational practice

Independent audit and public evidence

Secure code, configuration, evaluation, incident, appeal, deletion, and performance access for independent reviewers plus publication of non-sensitive findings.

KIT-PUB-PROC-16Operational practice

Exit, portability, remedy, and secure destruction

Require open export, migration support, continuity, deletion after exit, audit survival, damages or service credits, and rights that continue through merger or subcontractor change.

Common failure modes

Good intentions can still create cognitive surveillance.

KIT-PUB-FAIL-01Worldwide principle

The topic becomes the threat

A system treats a search about terrorism, protest, self-defense, religion, health, or government criticism as evidence about the person rather than evaluating conduct and context.

EvidenceCL-09-04
KIT-PUB-FAIL-02Technical recommendation

Administrative convenience becomes necessity

Full identity, bulk retention, or centralized access is adopted because it is easier to operate, not because less intrusive alternatives failed.

KIT-PUB-FAIL-03CognitiveLiberties.com policy proposal

Emergency becomes ordinary

An exceptional access or monitoring power loses its trigger, time limit, deletion duty, review, or budget endpoint and migrates into routine use.

KIT-PUB-FAIL-04Operational practice

Vendor secrecy replaces public accountability

A public body claims it cannot explain a decision because the model, score, data, or rules are proprietary.

EvidenceCL-15-02
KIT-PUB-FAIL-05Operational practice

Human review becomes ceremonial

The reviewer sees only the machine result, cannot inspect evidence, lacks authority to reverse it, or is measured for agreeing quickly.

EvidenceCL-07-04
KIT-PUB-FAIL-06Technical recommendation

Deletion removes the screen but not the inference

A visible record is erased while embeddings, labels, watch-list entries, training copies, and downstream consequences remain.

EvidenceCL-06-04
KIT-PUB-FAIL-07Jurisdiction-specific legal question

Public records become a pretext for exposure or erasure

Transparency is used to publish private inquiry, or privacy and classification are used to conceal official action and destroy accountability.

KIT-PUB-FAIL-08Worldwide principle

Government communication becomes unrecorded coercion

Officials use calls, meetings, threats, or procurement pressure to induce suppression while avoiding a reviewable order.

EvidenceCL-37-03
KIT-PUB-FAIL-09Technical recommendation

Aggregate reporting becomes re-identification

Small cells, rare categories, detailed geography, or repeated releases allow public statistics to expose individuals or protected communities.

EvidenceCL-10-02
KIT-PUB-FAIL-10Worldwide principle

A national rule is presented as a universal norm

A constitution, emergency doctrine, speech rule, records duty, or security policy from one jurisdiction is silently applied worldwide.

EvidenceCL-25-01
Incident-response procedure

Contain concrete harm without multiplying exposure.

  1. KIT-PUB-IR-01Operational practice

    Stabilize concrete and immediate harm

    Protect an identifiable person, stop an active fraud transaction, contain a system compromise, or preserve essential service while avoiding broader collection than the emergency requires.

    EvidenceCL-03-02
  2. KIT-PUB-IR-02Worldwide principle

    Classify evidence, authority, and urgency

    Record the observed facts, source reliability, legal authority, affected rights, time sensitivity, and what remains inference or allegation.

    EvidenceCL-15-02
  3. KIT-PUB-IR-03Technical recommendation

    Stop unnecessary collection and propagation

    Disable new scoring, syncing, sharing, model feedback, training ingestion, and downstream alerts that are not required for containment.

    EvidenceCL-10-01
  4. KIT-PUB-IR-04Operational practice

    Preserve the minimum accountable record

    Keep enough evidence to review official action and remedy harm while excluding unrelated queries, associations, communications, and bystanders.

  5. KIT-PUB-IR-05Operational practice

    Notify responsible and independent authorities

    Escalate to legal, privacy, security, records, accessibility, civil-rights, and independent-review owners according to the incident—not merely the system operator.

    EvidenceCL-25-04
  6. KIT-PUB-IR-06Worldwide principle

    Notify affected people when lawful and safe

    Give meaningful notice, the rule, evidence category, consequences, protection against retaliation, and a secure appeal route without repeating the exposure.

  7. KIT-PUB-IR-07Operational practice

    Require independent human reconsideration

    Review whether the trigger was accurate, the response was necessary and proportionate, the authority was valid, and less intrusive action was available.

  8. KIT-PUB-IR-08Worldwide principle

    Correct records and downstream decisions

    Withdraw unsupported labels, reverse avoidable adverse action, restore service or standing, notify downstream recipients, and preserve a correction history.

    EvidenceCL-07-04
  9. KIT-PUB-IR-09CognitiveLiberties.com policy proposal

    Delete data whose authority ended

    Remove emergency copies, temporary access, raw data, derived inferences, model feedback, and investigative material not subject to a valid continuing obligation.

  10. KIT-PUB-IR-10Operational practice

    Publish aggregate lessons and end the exceptional state

    Report cause, scope, errors, remedy, and control changes without exposing protected inquiry, then document that emergency powers, elevated access, and special retention have ended.

Review cadence

Make safeguards operational rather than ceremonial.

KIT-PUB-REV-01Operational practice

Monthly high-impact decision review

Sample adverse automated decisions, emergency access, reverse searches, government-platform communications, and sensitive-data exceptions for evidence, proportionality, appeal, and deletion.

EvidenceCL-15-02
KIT-PUB-REV-02Technical recommendation

Quarterly deletion and access review

Verify deletion jobs, backup expiry, privileged access, purpose separation, downstream correction, and closure of expired emergency authority.

EvidenceCL-10-01
KIT-PUB-REV-03Operational practice

Quarterly vendor and model-change review

Review subprocessors, policy changes, training use, evaluations, model drift, new inferences, accessibility, security, and contractual exceptions.

EvidenceCL-29-04
KIT-PUB-REV-04CognitiveLiberties.com policy proposal

Semiannual rights and impact review

Test representative political, religious, health, rights-related, minority-language, accessibility, and anonymous-use scenarios without retaining person-level inquiry histories.

EvidenceCL-15-01
KIT-PUB-REV-05Operational practice

Annual independent audit

Provide evidence to an independent body with authority to publish findings, order correction, and refer unlawful activity for appropriate review.

EvidenceCL-25-04
KIT-PUB-REV-06Operational practice

Event-driven review after legal or technical change

Reassess the system after law, case, emergency, model, vendor, data source, breach, merger, or mission change rather than waiting for the calendar.

EvidenceCL-21-04
KIT-PUB-REV-07CognitiveLiberties.com policy proposal

Sunset and dismantling review

Before renewal, require evidence of efficacy, error cost, necessity, alternatives, remaining harm, deletion, and a tested plan to terminate the power or system.

Appeals and remedy

A safeguard is incomplete when no one can reverse a mistake.

KIT-PUB-APP-01Worldwide principle

Visible human appeal route

Provide a channel that does not require the affected person to surrender additional unrelated cognitive, biometric, political, religious, or health data.

EvidenceCL-07-04
KIT-PUB-APP-02Operational practice

Meaningful reason and evidence category

State the rule, decision owner, machine role, material facts, confidence or uncertainty, and the consequence being imposed.

EvidenceCL-15-02
KIT-PUB-APP-03CognitiveLiberties.com policy proposal

Pause avoidable harm

Where safety permits, prevent benefit termination, license loss, discipline, account restriction, enforcement escalation, or public labeling until timely human review occurs.

EvidenceCL-07-04
KIT-PUB-APP-04Jurisdiction-specific legal question

Context, counsel, and accessible participation

Allow relevant documents, explanation, representation, interpretation, disability accommodation, language access, and a non-digital route.

EvidenceCL-42-04
KIT-PUB-APP-05Worldwide principle

Independent reviewer with reversal power

The appeal must not be decided solely by the original model, vendor, analyst, requester, or programme owner.

EvidenceCL-25-04
KIT-PUB-APP-06Operational practice

Defined decision deadlines

Publish urgency tiers and maximum time to acknowledge, review, correct, restore, and notify downstream systems.

EvidenceCL-07-04
KIT-PUB-APP-07Worldwide principle

Correction and downstream remedy

Repair the authoritative record, benefits, license, employment status, watch-list entry, public statement, platform request, and every known downstream copy.

EvidenceCL-20-03
KIT-PUB-APP-08Technical recommendation

Deletion of unsupported labels and features

Delete the invalid inference, score, embedding, feature, alert, training example, and future decision multiplier rather than merely closing the appeal ticket.

EvidenceCL-06-04
KIT-PUB-APP-09Jurisdiction-specific legal question

Institutional and individual remedy

Provide restoration, explanation, policy correction, fee or penalty reversal, compensation where authorized, disciplinary accountability, and public aggregate learning proportionate to the harm.

Data-deletion expectations

Delete the cognitive trail when the authorized need ends.

KIT-PUB-DEL-01Operational practice

Publish a record-by-record schedule

Name retention and deletion for prompts, searches, library activity, communications, location, biometrics, neural data, logs, scores, appeals, public records, legal holds, and backups.

EvidenceCL-06-04
KIT-PUB-DEL-02CognitiveLiberties.com policy proposal

Delete routine cognitive exhaust quickly

Remove query content, prompt text, reading events, detailed clickstreams, and session context as soon as the authorized transaction and security need end.

EvidenceCL-10-01
KIT-PUB-DEL-03Jurisdiction-specific legal question

Separate official records from private inquiry

Preserve accountable government decisions and orders without retaining unrelated intellectual activity merely because it passed through the same system.

EvidenceCL-20-03
KIT-PUB-DEL-04Technical recommendation

Expire emergency copies automatically

Emergency access, elevated privileges, temporary replicas, exports, and special logs must have technical expiration plus post-event review.

EvidenceCL-11-04
KIT-PUB-DEL-05Technical recommendation

Delete derived inferences with raw data

Removal includes risk labels, ideology or emotion estimates, relationship graphs, embeddings, vectors, model features, summaries, and cached outputs.

EvidenceCL-06-04
KIT-PUB-DEL-06Operational practice

Propagate correction and deletion

Notify and verify deletion across vendors, subprocessors, partner agencies, shared databases, backups, models, and public-facing systems where legally possible.

EvidenceCL-07-04
KIT-PUB-DEL-07Jurisdiction-specific legal question

Constrain legal holds and classified retention

A hold or classification must identify authority, custodian, scope, review date, release criteria, access roles, and why less material cannot be retained.

EvidenceCL-20-04
KIT-PUB-DEL-08Technical recommendation

Verify deletion rather than accepting attestation alone

Use deletion logs, sampled restoration tests, cryptographic erasure where appropriate, configuration evidence, and independent audit.

EvidenceCL-15-02
KIT-PUB-DEL-09Operational practice

Preserve correction provenance without preserving the dossier

Keep a minimal record that an official judgment was corrected and when, while removing the unsupported cognitive profile and unnecessary sensitive evidence.

EvidenceCL-20-03
KIT-PUB-DEL-10Operational practice

Terminate data custody at programme or vendor exit

Export lawful public records, migrate necessary service data, then securely destroy unnecessary government, vendor, backup, and model copies under verified closure.

Measurable outcomes without dossiers

Measure systems, controls, response, and recovery—not what named people think.

No metric in this kit requires an identity-linked history of lawful questions, reading, research, beliefs, associations, or use of privacy tools.

KIT-PUB-OUT-01Operational practice

Authority inventory coverage

Measure the percentage of high-impact systems, data demands, platform channels, emergency powers, and automated decisions with a named authority, owner, review path, and end condition.

Measurement boundary: Measure institutional documentation, not the beliefs, questions, or associations of individual people.

EvidenceCL-15-01
KIT-PUB-OUT-02Technical recommendation

Data minimization and separation

Measure eliminated fields, shortened retention, separated identities, purpose-bound stores, and privacy-preserving alternatives adopted.

Measurement boundary: Measure architecture and schemas rather than constructing a person-level history of lawful inquiry.

EvidenceCL-10-01
KIT-PUB-OUT-03Operational practice

Human-review coverage

Measure the share of material adverse decisions receiving qualified human review before final consequence and the reviewer’s actual reversal authority.

Measurement boundary: Measure decision process, not whether a person’s lawful interests match an institutional norm.

EvidenceCL-07-04
KIT-PUB-OUT-04Operational practice

Appeal timeliness and correction

Measure acknowledgment, decision, restoration, downstream correction, and deletion time by decision category using privacy-protective aggregates.

Measurement boundary: Use aggregate case timing without publishing or profiling the underlying questions or beliefs.

EvidenceCL-15-02
KIT-PUB-OUT-05Technical recommendation

Deletion assurance

Measure scheduled deletion completion, exception expiry, backup aging, derived-label removal, and independent verification.

Measurement boundary: Measure deletion jobs and exceptions rather than retaining cognitive data to prove deletion later.

EvidenceCL-06-04
KIT-PUB-OUT-06Operational practice

Government-platform transparency

Measure recorded communications, legal-basis completeness, response categories, rejected requests, appeal availability, and publication lag.

Measurement boundary: Report aggregate official action without exposing lawful speakers, readers, sources, or complainants.

EvidenceCL-37-04
KIT-PUB-OUT-07CognitiveLiberties.com policy proposal

Emergency-power closure

Measure expiration, elevated-access removal, temporary-copy deletion, reauthorization evidence, and dismantling-test completion.

Measurement boundary: Measure whether exceptional machinery ended, not the private activity of the population subjected to it.

EvidenceCL-11-04
KIT-PUB-OUT-08Technical recommendation

Synthetic false-positive performance

Use controlled synthetic scenarios to test sensitive topics, minority languages, disability, rights research, anonymous access, and context without enrolling real people into risk profiles.

Measurement boundary: Use synthetic or consented testing rather than longitudinal surveillance of ordinary service users.

EvidenceCL-15-01
KIT-PUB-OUT-09Operational practice

Procurement and vendor accountability

Measure contract coverage, subprocessor visibility, audit access, model-change notice, training restrictions, incident notification, export, remedy, and verified exit.

Measurement boundary: Measure contracts and vendor evidence, not personal thought histories.

EvidenceCL-29-04
KIT-PUB-OUT-10Operational practice

Public-record integrity and access

Measure fixity, provenance, correction linkage, declassification review, accessible formats, restoration, and independent archive custody.

Measurement boundary: Measure record-system integrity without exposing protected private inquiry or confidential sources.

EvidenceCL-20-03
KIT-PUB-OUT-11Operational practice

Oversight independence and completion

Measure audit scope, evidence access, unresolved findings, corrective deadlines, whistleblower protection, and closure verified by an independent body.

Measurement boundary: Measure oversight power and remediation rather than scoring the ideology or trustworthiness of individuals.

EvidenceCL-25-04
KIT-PUB-OUT-12CognitiveLiberties.com policy proposal

Anonymous and low-data access availability

Measure the share of public information and ordinary services available without full identity, plus accessibility and documentation alternatives.

Measurement boundary: Measure service design and access routes rather than tracking who chooses anonymity or why.

EvidenceCL-17-04
Jurisdiction-specific legal review

Ask these questions locally before claiming compliance.

The worldwide baseline is a rights and architecture framework. Binding duties vary across constitutions, human-rights systems, privacy, consumer, education, press, labor, accessibility, cybersecurity, records, procurement, contracts, and court procedure.

Evidence basis and limits

Research, standards, law, technical guidance, and site proposals remain distinguishable.

This kit translates supplied research, human-rights materials, selected court and oversight records, technical standards, and CognitiveLiberties.com policy into a worldwide implementation programme. It does not declare one jurisdiction’s law universal, certify legal compliance, authorize surveillance, or substitute for local constitutional, criminal-procedure, public-records, procurement, accessibility, labor, education, benefits, national-security, and emergency-law review.

Current law, vendor behavior, system configuration, and local risk must be independently rechecked before deployment. The exact 64-report archive remains preserved separately from this implementation derivative.