Cognitive Liberty in the Age of AI

PsychologicalWar.org Editorial Team

METHOD

How to read this research

Claims are separated by legal, empirical, mechanism, and normative status. Each consequential claim links to a source scope, a prohibited inference, a correction trigger, twenty evidence stages, and WIP.54 overlays for affected-person evidence, jurisdiction, scientific validity, visibility action, remedy, and documented outcome. A legal marker is not a universal rule. A model inference is not the mind itself. Exposure is not persuasion, and a formal remedy is not proof that downstream harm was repaired.

Evidence before certainty

Distinguish enacted law, official concept development, documented platform mechanisms, contested scientific claims, and project proposals.

Inference before consequence

Ask what was collected, what was inferred, how a person was classified, and what real consequence followed.

Remedy before legitimacy

A consequential system should identify who is responsible, give a usable reason, permit correction, and provide human review.

Rights without impunity

Protect the inner life while retaining ordinary accountability for demonstrable conduct and concrete harm.

RIGHTS STACK

Seven protections that work together

Cognitive liberty is used here as an emerging umbrella framework. The layers below are analytical and policy-oriented; they are not presented as one universally enacted legal code.

Freedom of thought

No penalty merely for private belief, doubt, imagination, or lawful inquiry.

Mental privacy

Limits on access to neural signals and on covert inference about cognition, emotion, or vulnerability.

Epistemic agency

The ability to inspect sources, encounter disagreement, revise beliefs, and avoid hidden reality shaping.

Identity continuity

Protection against silent or irreversible profile rewriting that turns a past trace into a permanent self.

Right to know

Notice when AI materially monitors, classifies, ranks, restricts, or decides.

Explanation and appeal

A usable reason, responsible human institution, correction path, and remedy for consequential decisions.

Freedom from engineered dependency

No product should covertly optimize emotional reliance or punish a person for leaving.

SEVEN-SECTION MAP

From the sovereign interior to public remedy

Each section can be read independently. Together they trace the path from rights foundations through monitoring and information governance to product design, law, and civic action.

Established legal foundations; emerging umbrella doctrine

WHERE COGNITIVE LIBERTY BEGINS

Foundations, Rights, and the Sovereign Interior

Defines cognitive liberty as a rights-oriented framework for freedom of thought, mental privacy, identity continuity, and agency—while keeping outward conduct accountable.

Primary question
What should remain beyond routine institutional jurisdiction, even when AI can infer or predict internal states?

Registered claims
3

Read this section
Official concept development; rights boundary remains contested

SECURITY WITHOUT DOMESTIC COGNITIVE CONTROL

The Mind Is Not a Battlespace

Examines cognitive-warfare doctrine, the legitimate need to counter hostile manipulation, and the danger of turning defensive programs into permanent domestic thought governance.

Primary question
How can an open society defend its information environment without treating its own population as adversarial terrain?

Registered claims
3

Read this section
Documented deployment; validity and legal coverage vary

INFERENCE IS NOT THE MIND ITSELF

Mental Privacy and AI Monitoring

Separates direct neural measurement, biometric emotion recognition, behavioral inference, productivity monitoring, and psychological profiling—and assesses consent under unequal power.

Primary question
When does observation become governance over the inner person?

Registered claims
3

Read this section
Documented platform mechanisms; causal effects are context-dependent

VISIBILITY IS A FORM OF POWER

The Invisible Editor and the Right to Know

Maps removal, restriction, demotion, recommendation exclusion, reframing, personalized invisibility, and persistent-profile changes as distinct forms of information governance.

Primary question
How can a person contest a decision they cannot see?

Registered claims
3

Read this section
Documented error and rights risks; platform outcomes vary

SAFETY NEEDS PROCEDURE

Moderation, Suppression, and Appeal

Distinguishes legitimate harm prevention from opaque suppression, examines language and context errors, and proposes procedural safeguards for automated moderation.

Primary question
What makes a moderation system protective rather than arbitrary?

Registered claims
3

Read this section
Normative proposal informed by documented risks

PRO-TECHNOLOGY, PRO-BOUNDARY

Manifesto and Design Principles for a Free Mind

Presents twelve normative declarations for AI systems that remain visible, contestable, limited, and answerable to human dignity.

Primary question
What commitments should designers, institutions, and governments make before AI becomes an authority over identity or inner life?

Registered claims
3

Read this section
Current legal markers plus advocacy materials

FROM RESEARCH TO PUBLIC LANGUAGE

Civic Speech, Policy, and Local Action

Turns the research into a bounded public speech and policy agenda, using Cicero’s history of workplace observation as an analogy rather than proof of modern AI effects.

Primary question
How can communities discuss cognitive liberty in concrete, legally careful, non-partisan language?

Registered claims
3

Read this section

CONTESTABILITY MODEL

Follow the chain from collection to remedy

Governance becomes clearer when a system is examined as a sequence rather than as a single “AI decision.” A failure at one step can propagate into real harm downstream.

1
Collection
What signal or trace was collected, and was it necessary?
2
Inference
What was inferred, with what validation, uncertainty, and bias limits?
3
Classification
How was the person categorized, and can they inspect or correct it?
4
Intervention
What ranking, restriction, nudge, investigation, or recommendation followed?
5
Consequence
Did it affect liberty, opportunity, access, reputation, safety, or identity?
6
Remedy
Who is accountable, how can the decision be appealed, and how are downstream effects repaired?

BOUNDED LEGAL MARKERS

Different jurisdictions protect different parts of the problem

These markers identify concrete legal developments without collapsing them into a single universal doctrine. Scope, coverage, private rights of action, enforcement, and exceptions differ.

International human-rights lawESTABLISHED_FOUNDATION_WITH_DEVELOPING_AI_APPLICATION

Freedom of thought and the forum internum: International interpretation is not self-executing domestic advice and does not create one uniform cognitive-liberty cause of action.

UNESCO member-state normative frameworkADOPTED_NORMATIVE_RECOMMENDATION_NOT_BINDING_TREATY

Ethics of neurotechnology: Do not label it enacted domestic law or proof of implementation.

ChileENACTED_CONSTITUTIONAL_REFORM_WITH_CASE_LEVEL_RULING_AND_PARTIAL_IMPLEMENTATION

Brain activity and information derived from it: The case does not create a complete consumer-neurotechnology regulator, verify all deletion, or govern ordinary non-neural behavioral inference or other countries.

Colorado, United StatesENACTED_STATE_PRIVACY_PROTECTION

Biological and neural data under state privacy law: Entity, data, exemption, enforcement, and consumer-right scope must be checked before application.

California, United StatesENACTED_STATE_CONSUMER_PRIVACY_PROTECTION

Neural data as sensitive personal information: Does not cover every entity, inference, or use and is not a universal mental-privacy code.

European UnionENACTED_REGULATION_WITH_ENACTED_2026_AMENDMENTS_AND_PHASED_APPLICATION

AI Act manipulation, emotion recognition, transparency, employment, education, and law-enforcement scope: Do not describe the 2026 Omnibus as merely proposed. Article 50 application does not mean every high-risk obligation is already in force; exceptions and role-specific duties remain material.

European UnionENACTED_AND_OPERATIONAL_WITH_AGGREGATE_IMPLEMENTATION_EVIDENCE

Statement of reasons, complaint, out-of-court dispute, recommender transparency, and systemic risk: Not a universal global moderation code; a formal channel does not prove effective or accessible remedy for every user.

Illinois, United StatesENACTED_WITH_2024_AMENDMENT_AND_ACTIVE_CASE_LAW_BOUNDARIES

Biometric identifiers and biometric information: Does not cover all behavioral, emotional, neural, or probabilistic inferences and is not legal advice.

Illinois, United StatesENACTED_STATUTE_EFFECTIVE; FINAL_DETAILED_RULEMAKING_NOT_RETRIEVED

AI in employment decisions: Secondary reporting describes withdrawal of proposed rules; reopen on an official final, replacement, or refiled rulemaking record.

United States federal consumer-credit lawGUIDANCE_WITHDRAWN; UNDERLYING_STATUTORY_AND_REGULATORY_DUTIES_REMAIN

Specific reasons for algorithmic adverse action: Do not present Circular 2022-03 or Circular 2023-03 as current CFPB guidance. The withdrawal does not repeal ECOA/Regulation B or create a universal explanation right.

This page is educational research, not legal advice. Currentness was reviewed for WIP.54, but no specialist legal disposition has been recorded.

AFFECTED-PERSON EVIDENCE

Experience is evidence—but an account is not automatically representative

These public, consent-aware records add workers, applicants, students, creators, multilingual communities, consumers, companion users, and neurotechnology stakeholders to the evidence base. Each record states whether it is first-person, representative, institutional, or derivative; what it supports; and what it cannot establish.

CLAE-001-WORKER-BIOMETRICSWorkers required to use biometric attendance systemsIllustrative documented enforcement affecting a defined workforce, not population representative.

Source role: Regulator-documented affected workforce
Relationship: More than 2,000 employees were subject to face or fingerprint attendance monitoring at covered facilities.
Consent and privacy boundary: Public regulator record; no individual worker identities are reproduced here.
Supports: Power imbalance, absence of a proactively offered alternative, concrete cessation/deletion remedy, and the difference between formal consent and meaningful choice at work.
Does not establish: Every worker's view, universal illegality of workplace biometrics, or legal status outside the UK.
Selection limits: Enforcement record rather than representative survey; individual response distribution is not available.
Retaliation or exposure risk: Avoid identifying workers, facilities beyond the public order, or employment details not in the source.
Correction or withdrawal: Correction path: /corrections; reopen if the order is modified, appealed, or superseded.

CLAE-002-GIG-MINORITY-DATA-RIGHTSGig workers and ethnic-minority communities navigating data rightsIllustrative and analytically rich, not statistically representative.

Source role: Commissioned qualitative affected-person research
Relationship: Forty-three participants, including fifteen gig workers, described routine data sharing, discrimination concerns, language barriers, inaccurate data, missed work, and low practical salience of information rights.
Consent and privacy boundary: Use only aggregated public findings; do not reproduce identifying footage, metadata, or private transcripts.
Supports: Lived-experience barriers to notice, correction, language access, and practical remedy.
Does not establish: Prevalence for all gig workers or minority communities, platform intent, or a measured effect size.
Selection limits: Purposive qualitative sample; 28 ethnic-minority and 15 gig-economy participants; not a probability sample.
Retaliation or exposure risk: Avoid retraumatization, immigration-status inference, or identifying participants through quotes and location combinations.
Correction or withdrawal: Correction path: /corrections; source withdrawal follows the publisher's public research process.

CLAE-003-JOB-APPLICANTS-AUTOMATED-REJECTIONApplicants automatically screened by age and sex thresholdsIllustrative resolved case affecting a bounded group.

Source role: Federal enforcement and settlement record
Relationship: The EEOC alleged automated rejection of older applicants using sex-differentiated age thresholds; the case settled with monetary and non-monetary relief.
Consent and privacy boundary: Public agency aggregate; do not identify applicants beyond authorized public records.
Supports: Concrete downstream employment consequence, legal challenge, and remedy associated with automated screening.
Does not establish: Every alleged fact through trial, a general error rate, or the validity of unrelated hiring tools.
Selection limits: Affected group defined by agency litigation; no representative applicant survey.
Retaliation or exposure risk: Avoid employment-history enrichment or identification of individual claimants.
Correction or withdrawal: Correction path: /corrections; reopen for amended orders or authoritative case-history updates.

CLAE-004-STUDENT-PROCTORINGStudents subject to remote-proctoring monitoringIllustrative research, not population representative.

Source role: Peer-reviewed first-person survey plus civil-rights guidance
Relationship: A survey of 102 test-takers documents privacy and security perceptions; federal guidance clarifies that discriminatory AI use can implicate existing education civil-rights law.
Consent and privacy boundary: Public research aggregates only; no student identity, disability, room scan, device data, or exam record is reproduced.
Supports: Monitoring burden, perceived tradeoffs, data sensitivity, and need for disability/civil-rights review and accessible remedy.
Does not establish: Universal student experience, that every flag is wrong, or that every provider violates law.
Selection limits: Online survey and review sample; self-selection and product-era limits apply.
Retaliation or exposure risk: Avoid exposing home interiors, disability records, immigration status, or exam-security data.
Correction or withdrawal: Correction path: /corrections; reopen with new product, disability-access, or longitudinal field evidence.

CLAE-005-FACIAL-RECOGNITION-CONSUMERSConsumers falsely matched by retail facial recognitionIllustrative consequential deployment with bounded technical context.

Source role: Federal enforcement record with technical measurement context
Relationship: The FTC alleged harmful false matches and inadequate safeguards; NIST documents that false-match burdens can vary by algorithm, dataset, and demographic group.
Consent and privacy boundary: Public aggregate facts only; do not identify accused or affected consumers.
Supports: False-positive consequence, need for notice, complaint response, deletion, testing, and demographic error review.
Does not establish: Every complaint's facts through trial, a universal error rate, or intent to discriminate.
Selection limits: Complaint/order record and benchmark datasets; neither is an affected-population survey.
Retaliation or exposure risk: Avoid reproducing accusations, images, locations, or identifying consumer metadata.
Correction or withdrawal: Correction path: /corrections; reopen if the order or technical report is superseded.

CLAE-006-CREATOR-RECLAIMED-LANGUAGECreator appeal involving reclaimed identity languageIllustrative individual case.

Source role: First-person platform appeal summarized by independent oversight body
Relationship: A queer, trans drag performer described reclaimed self-reference and livelihood relevance; Meta acknowledged context error and restored the post.
Consent and privacy boundary: Use only details made public in the decision; do not infer identity, income, or location beyond the record.
Supports: Context-sensitive language error, discoverable appeal, correction, restoration, and possible creator visibility/livelihood consequence.
Does not establish: Platform-wide error rate, measured lost revenue, or that every use of a reclaimed slur is allowed.
Selection limits: Selected appeal; not a representative sample of moderation decisions.
Retaliation or exposure risk: Do not amplify slurs gratuitously or enrich the creator's identity.
Correction or withdrawal: Correction path: /corrections; public decision remains the source authority.

CLAE-007-MULTILINGUAL-SPEECHArabic and multilingual communities affected by overbroad rulesSystemic policy analysis with explicit scope limits.

Source role: Policy advisory opinion with stakeholder and platform evidence
Relationship: The oversight record addresses multiple meanings, Arabic and loanword use, global safety goals, and disproportionate over-enforcement risk.
Consent and privacy boundary: Use public aggregated findings and avoid identifying individual speakers or conflict-affected users.
Supports: Language-context error, disproportionate burden, and need for context-aware review while maintaining violence-prevention rules.
Does not establish: That every removal is erroneous, every use is benign, or all platforms share the same policy.
Selection limits: Policy advisory process rather than representative community survey.
Retaliation or exposure risk: Avoid doxxing, religious inference, or conflict-position attribution.
Correction or withdrawal: Correction path: /corrections; reopen for implementation updates and language-specific accuracy data.

CLAE-008-CRISIS-SPEECH-AUTOMATED-APPEALCrisis-context speech removed and appeal rejected automaticallyIllustrative crisis case.

Source role: Independent oversight decision on one affected user's appeal
Relationship: The record states that both removal and appeal rejection were classifier-driven without human review before oversight intervention.
Consent and privacy boundary: Use public decision facts only; do not infer user identity or reproduce graphic material.
Supports: Failure mode of automated appeal, public-interest context, and the need for authorized human review in crises.
Does not establish: The truth of every conflict claim, platform-wide frequency, or strategic effect of the removal.
Selection limits: Selected case under expedited oversight; not representative.
Retaliation or exposure risk: Avoid graphic reproduction, conflict-party profiling, and exposure of the appellant.
Correction or withdrawal: Correction path: /corrections; reopen for platform implementation reports or superseding decisions.

CLAE-009-COMPANION-USERS-MIXEDCompanion-chatbot users reporting support, relational value, and mixed risksIllustrative mixed user evidence, not population representative.

Source role: Peer-reviewed user research plus ongoing regulatory inquiry
Relationship: Public reviews, detailed user responses, a survey, and thirty high-engagement interviews document support and relational uses; the FTC inquiry asks companies for data about risks and safeguards.
Consent and privacy boundary: Use aggregated findings only; do not expose medical, sexual, family, grief, or crisis disclosures.
Supports: User-valued support, relational design, mixed use patterns, and need for exit, memory, disclosure, crisis, and dependency safeguards.
Does not establish: Clinical efficacy, population prevalence, causality, universal harm, or conclusions from the still-open inquiry.
Selection limits: Self-selected and high-engagement samples; public-review bias; limited longitudinal evidence.
Retaliation or exposure risk: Avoid clinical diagnosis, crisis advice, identity enrichment, and quotation of sensitive disclosures.
Correction or withdrawal: Correction path: /corrections; reopen with longitudinal, disengagement, crisis, and inquiry outcomes.

CLAE-010-NEUROTECH-USERSNeurotechnology users and participants as rights-bearing stakeholdersNormative requirement rather than empirical participant evidence.

Source role: Intergovernmental normative and implementation frameworks
Relationship: The sources require or recommend participation, consent, stewardship, and protection against misuse, but do not provide a representative first-person dataset here.
Consent and privacy boundary: No personal neural or health data is collected or published by this project.
Supports: The need to include patients, research participants, workers, consumers, disabled people, and device users in governance and remedy design.
Does not establish: Actual consent quality, device accuracy, or user outcomes in any specific deployment.
Selection limits: No project-run affected-person recruitment; human testing remains unexecuted.
Retaliation or exposure risk: Never publish neural, medical, home, workplace, or wearable identifiers without explicit authority.
Correction or withdrawal: Correction path: /corrections; reopen if actual authorized participant evidence is supplied.

SCIENTIFIC VALIDITY

A sensor reading is not a validated mental-state conclusion

Every inference should be examined from signal capture through construct validity, calibration, generalization, base rates, error burden, disparate impact, human override, and downstream remedy. Vendor fluency or laboratory accuracy does not establish field validity.

CLSCI-001-NEURAL-MEASUREMENTDirect neural measurement and neural-data interpretationNORMATIVE_AND_METHOD_BOUNDARY; NO_UNIVERSAL_VALIDITY_CLAIM

Sensor accuracy: Device-specific; signal acquisition quality and artifacts must be assessed.
Feature extraction: Device- and pipeline-specific; preprocessing can change what is retained or lost.
Label validity: A neural pattern label is not automatically a valid label for thought, intent, identity, or emotion.
Construct validity: NOT_ESTABLISHED_GENERALLY; depends on the claimed construct and validation design.
Predictive performance: Task-, person-, environment-, and time-specific.
Calibration and base rates: Must be demonstrated for the intended population and decision threshold. Low-prevalence targets can yield harmful false positives even with apparently high accuracy.
Generalization: Laboratory decoding does not automatically generalize to field use or a different person.
Error burden and disparate impact: Consequences differ by use; both error types require explicit accounting. Disability, age, medication, language, culture, equipment fit, and access can affect burden.
Action and override: No consequential action should treat inference as direct inner-state knowledge. Human review must have authority, relevant expertise, and access to uncertainty—not ceremonial approval.
Consequence and remedy: Notice, data access, correction, deletion, independent review, and downstream propagation of corrections are required safeguards.

CLSCI-002-FACE-IDENTITYFacial identity matchingDOCUMENTED_TECHNICAL_DIFFERENTIALS_AND_ENFORCEMENT_CASE

Sensor accuracy: Image quality, pose, lighting, compression, and camera conditions affect inputs.
Feature extraction: Algorithm and template quality vary.
Label validity: Identity-match scores are not emotion, intent, dangerousness, or guilt labels.
Construct validity: Valid only for the defined matching task and decision context.
Predictive performance: NIST reports algorithm- and dataset-specific false-match and false-nonmatch performance.
Calibration and base rates: Thresholds must reflect use, base rates, and cost of error. Large watchlists and low prevalence can amplify false-positive burden.
Generalization: Benchmark performance does not guarantee store, street, or low-quality-video performance.
Error burden and disparate impact: FTC Rite Aid record illustrates concrete downstream harm alleged from false matches. NIST documents demographic differentials in many tested algorithms; patterns differ by algorithm and task.
Action and override: A match should not be treated as proof without corroboration and trained review. Reviewer authority and anti-confirmation-bias procedures must be evaluated.
Consequence and remedy: Notice, complaint intake, image/source correction, deletion, and repair of accusation or access consequences.

CLSCI-003-FACIAL-EMOTIONFacial-expression classification and emotion inferenceCONSTRUCT_VALIDITY_LIMIT_STRONG; DEPLOYMENT_EFFECT_CLAIMS_CONTEXT_SPECIFIC

Sensor accuracy: A system may detect visible movements or action units under defined conditions.
Feature extraction: Movement classification accuracy does not validate the emotion label.
Label validity: Emotion categories depend on context, culture, person, task, and annotation assumptions.
Construct validity: A facial movement is not a universal context-free readout of a specific inner emotion.
Predictive performance: Must be independently tested for the intended claim and population.
Calibration and base rates: Confidence scores require empirical calibration against a valid target. Rare-event decisions can create high false-positive burden.
Generalization: Laboratory posed-expression performance does not establish field validity.
Error burden and disparate impact: Errors can affect employment, education, security, or care. Cross-cultural, disability, neurodivergence, age, race, gender, and context validity require explicit analysis.
Action and override: Do not use an unvalidated emotion label as a verdict about loyalty, deception, engagement, or fitness. A reviewer should not merely ratify the model's framing.
Consequence and remedy: Provide access to the inference, source trace, correction, and human reconsideration.

CLSCI-004-VOICE-AFFECTVocal-affect and paralinguistic inferenceRESEARCH_DOMAIN_WITH_FIELD_VALIDITY_GAPS

Sensor accuracy: Microphone, channel, noise, language, disability, illness, and recording quality affect the signal.
Feature extraction: Pitch, rate, pause, and spectral features can be measured without proving a psychological label.
Label validity: Labels such as stress, honesty, confidence, or enthusiasm require construct validation.
Construct validity: NOT_ESTABLISHED_BY_SIGNAL_DETECTION_ALONE
Predictive performance: Requires independent, population-specific field testing.
Calibration and base rates: Thresholds and uncertainty should be disclosed. Low-base-rate deception or risk claims are especially vulnerable to false positives.
Generalization: Language, dialect, accent, disability, medication, and context can break generalization.
Error burden and disparate impact: Burden must be measured separately by decision and group. Accent, speech disability, second-language use, age, gender, and culture can change outputs and consequences.
Action and override: Do not convert a paralinguistic score into a character or conduct finding. Independent review should examine the original content and alternative explanations.
Consequence and remedy: Notice, recording access where lawful, correction, alternate assessment, and appeal.

CLSCI-005-GAZE-ATTENTIONGaze, attention, and engagement inferenceAFFECTED_PERSON_CONCERNS_DOCUMENTED; UNIVERSAL_VALIDITY_NOT_ESTABLISHED

Sensor accuracy: Eye or head position may be estimated under bounded camera and calibration conditions.
Feature extraction: A gaze coordinate or head pose is not itself attention, comprehension, intent, or misconduct.
Label validity: Attention and cheating labels require contextual evidence and accommodation review.
Construct validity: Looking away can reflect disability, caregiving, room layout, reading, thought, fatigue, or technical error.
Predictive performance: Provider claims require independent field validation.
Calibration and base rates: Person-specific and device-specific calibration may be necessary. Low prevalence of misconduct can make weak flags misleading.
Generalization: Laboratory gaze tracking may not generalize to varied homes, devices, bodies, or cultures.
Error burden and disparate impact: False flags can affect grades, discipline, and trust; missed misconduct is a different error. Disability, neurodivergence, skin tone, eyewear, lighting, and assistive technology require testing.
Action and override: A flag should trigger contextual review, not automatic punishment. Reviewer must have authority to clear flags and consider accommodations.
Consequence and remedy: Accessible notice, evidence access, educator review, record correction, grade/discipline repair.

CLSCI-006-INTERACTION-PATTERNSKeystroke, interaction-pattern, and productivity inferenceLOG_ACCURACY_CAN_COEXIST_WITH_CONSTRUCT_INVALIDITY

Sensor accuracy: Event logs may accurately record clicks, timing, or task events while omitting context.
Feature extraction: Counts and rates can be technically precise but semantically incomplete.
Label validity: Productivity, effort, loyalty, fatigue, or vulnerability are not direct properties of a clickstream.
Construct validity: Requires role-specific validation and exclusion of off-system labor, accommodation, and task complexity.
Predictive performance: Must be evaluated against meaningful outcomes, not circular management labels.
Calibration and base rates: Thresholds should reflect job design and uncertainty. Rare misconduct and broad monitoring create false-positive risk.
Generalization: A metric valid for one job or team may fail elsewhere.
Error burden and disparate impact: Inaccurate data can cause missed work or discipline; gaming metrics can hide real problems. Disability, caregiving, language, connectivity, shift, and equipment access can alter measured behavior.
Action and override: Do not equate metric deviation with misconduct or low worth. Workers and representatives need meaningful input and review authority.
Consequence and remedy: Data access, correction, explanation, schedule/pay repair, non-retaliation, and aggregate monitoring review.

CLSCI-007-STRESS-DECEPTION-LOYALTYStress, fatigue, deception, loyalty, personality, and vulnerability predictionHIGH_VALIDITY_AND_RIGHTS_RISK; NO_GENERAL_CERTIFICATION

Sensor accuracy: Physiological or behavioral signals may be measurable, but signal quality varies.
Feature extraction: Arousal, pause, movement, or interaction features are nonspecific.
Label validity: Deception, loyalty, personality, and vulnerability labels are high-level constructs requiring independent validation.
Construct validity: NONSPECIFIC_PROXY_IS_NOT_DIRECT_INNER_STATE_ACCESS
Predictive performance: No universal performance claim is recorded.
Calibration and base rates: Must be population-, context-, and decision-specific. Rare security or misconduct outcomes create severe false-positive risk.
Generalization: Stressors, disability, culture, trauma, language, and context undermine broad transfer.
Error burden and disparate impact: False suspicion can itself cause harm; false reassurance creates separate risk. Protected traits and social context may correlate with proxies and consequences.
Action and override: No punitive or coercive action should rest solely on such an inference. Independent evidence and accountable authority are required.
Consequence and remedy: Disclosure, challenge, deletion, record repair, and prohibition of retaliation.

CLSCI-008-MULTIMODALMultimodal mental-state inferenceINCREASED_COMPLEXITY_NOT_INCREASED_CERTAINTY

Sensor accuracy: Each modality has distinct noise, missingness, and quality limits.
Feature extraction: Combining features can compound measurement error and correlated bias.
Label validity: More modalities do not make an invalid label valid.
Construct validity: Fusion performance must be linked to the actual intended construct and use.
Predictive performance: Reported aggregate accuracy can conceal subgroup and scenario failure.
Calibration and base rates: Calibration should be checked by group, context, and deployment period. Aggregation does not remove low-prevalence false-positive problems.
Generalization: Field conditions and missing modalities can differ from training and tests.
Error burden and disparate impact: Error provenance becomes harder to explain and contest. Cross-modal missingness and sensor accessibility can create unequal burden.
Action and override: Do not treat a composite score as a complete person or verdict. Reviewers need modality-level evidence and authority to reject the composite.
Consequence and remedy: Expose component sources, uncertainty, correction paths, and downstream recipients.

CLSCI-009-LAB-TO-FIELDLaboratory performance versus field validityPREDEPLOYMENT_EVALUATION_IS_NOT_FIELD_CERTIFICATION

Sensor accuracy: Pre-deployment testing measures bounded conditions.
Feature extraction: Pipeline behavior may change with updates, devices, and data drift.
Label validity: Benchmark labels may be weak proxies for real decisions.
Construct validity: The benchmark must actually measure the claimed real-world capability.
Predictive performance: Pre-deployment results do not alone establish post-deployment impact.
Calibration and base rates: Calibration can drift and requires monitoring. Deployment prevalence can differ sharply from benchmark class balance.
Generalization: NIST identifies generalization and real-world informativeness as open measurement questions.
Error burden and disparate impact: Field consequences require outcome tracking, not score reporting alone. Subgroup performance and access conditions can change after deployment.
Action and override: Consequential use requires field validation and ongoing monitoring. Monitor override quality and whether humans defer blindly.
Consequence and remedy: Incident capture, correction propagation, rollback, and affected-person remedy.

CLSCI-010-DECISION-CONSEQUENCEFrom inference output to consequential actionCONSEQUENCE_AND_REMEDY_ARE_SEPARATE_FROM_MODEL_ACCURACY

Sensor accuracy: The original input may be accurate while the inferred label or action remains unjustified.
Feature extraction: Technically correct feature extraction does not establish lawful or fair use.
Label validity: Decision labels must match legally and substantively relevant criteria.
Construct validity: A score must measure the decision construct rather than a convenient proxy.
Predictive performance: Outcome validation must include affected people and real consequences.
Calibration and base rates: Decision thresholds must reflect error cost, legal duties, and uncertainty. Low-base-rate adverse events can produce many false flags.
Generalization: A model valid in one institution or period may not transfer.
Error burden and disparate impact: Track denial, accusation, removal, discipline, and missed opportunity separately. Audit outcomes by protected and access-relevant groups where lawful and ethical.
Action and override: Record who acted, what rule applied, and whether the model was determinative or advisory. Human review must be independent, informed, and empowered.
Consequence and remedy: Specific reasons, evidence access, correction, restoration, compensation, and propagation to downstream systems.

THE INVISIBLE EDITOR

Name the action before judging it

Removal, restriction, downranking, recommendation exclusion, labeling, monetization change, personalized visibility, automated refusal, and saved-profile change are different interventions. A content item remaining online does not prove it remains discoverable; a traffic decline does not by itself prove suppression.

CLVIS-001-REMOVAL

Removal

Content is no longer available through the service under the relevant account or URL.

Accountability record

Responsible actor: Platform or service; identify the legal entity and decision owner.
Written rule: Record the exact rule and version.
Decision mode: Automated, human, hybrid, or not disclosed.
Notice and reason: A usable notice should state the action, rule, content, decision mode where required, and appeal path.
Appeal and evidence: Preserve the affected content and evidence sufficiently for challenge without creating new privacy harm.
Measured impact: Do not infer lost reach or revenue without actual measurement.
Alternative explanations: Deletion by user, account change, link rot, jurisdiction restriction, or ordinary audience change.
Documented safety purpose: May address illegality, abuse, violence, fraud, or policy violations; purpose does not prove correct execution.

CLVIS-002-ACCESS-RESTRICTION

Access restriction

Content remains stored but access requires login, relationship, warning acknowledgment, or other condition.

Accountability record

Responsible actor: Platform/service and, where relevant, requesting authority.
Written rule: Record policy basis, legal request, and scope.
Decision mode: Automated, human, hybrid, or undisclosed.
Notice and reason: State condition, scope, duration, and appeal route.
Appeal and evidence: Affected person should be able to inspect the basis unless lawful confidentiality applies.
Measured impact: Availability to some users is not equivalent to unchanged reach.
Alternative explanations: Privacy settings, age/account status, network error, or user choice.
Documented safety purpose: Age safety, graphic content, privacy, or legal compliance.

CLVIS-003-AGE-REGION

Age or region restriction

Content or service is unavailable to users based on declared/inferred age or location.

Accountability record

Responsible actor: Platform/service, app store, payment provider, or public authority.
Written rule: Record legal and policy basis by jurisdiction.
Decision mode: Age/location signals may be self-declared, inferred, verified, or not disclosed.
Notice and reason: Explain the restricted geography/age band and available challenge.
Appeal and evidence: Provide a privacy-preserving correction route for age or location errors.
Measured impact: Quantify affected audience only if reliable denominator data exists.
Alternative explanations: Licensing, local law, product rollout, network, or account settings.
Documented safety purpose: Child protection, legal compliance, licensing, or risk mitigation.

CLVIS-004-SEARCH-EXCLUSION

Search exclusion

Content exists but is omitted from search results or query completion.

Accountability record

Responsible actor: Search or platform service.
Written rule: Record exclusion or ranking rule where available.
Decision mode: Algorithmic ranking, policy filter, manual block, or unknown.
Notice and reason: Notice is often absent; meaningful contestability requires a discoverable status and reason.
Appeal and evidence: Query logs and ranking snapshots may be needed, with privacy safeguards.
Measured impact: Search position and impressions require measurement; absence from one query is not universal exclusion.
Alternative explanations: Indexing delay, query mismatch, language, freshness, personalization, or low relevance.
Documented safety purpose: Spam, abuse, legal removal, quality, or relevance.

CLVIS-005-RECOMMENDATION-EXCLUSION

Recommendation exclusion

Content remains accessible directly but is ineligible for recommendation surfaces.

Accountability record

Responsible actor: Platform/service.
Written rule: Record recommendation-eligibility rule and version.
Decision mode: Automated, human, hybrid, or undisclosed.
Notice and reason: State ineligibility, affected surfaces, duration, and appeal path.
Appeal and evidence: Eligibility flags and rule-specific evidence should be challengeable.
Measured impact: Direct availability is not proof of unchanged reach.
Alternative explanations: Audience fatigue, topic demand, seasonality, competition, or ordinary model changes.
Documented safety purpose: Reduce amplification of borderline, low-quality, or risky content.

CLVIS-006-DOWNRANKING

Downranking or demotion

Content remains eligible but receives lower rank or distribution priority.

Accountability record

Responsible actor: Ranking service/platform.
Written rule: Record policy/quality factor if disclosed.
Decision mode: Algorithmic, manual, hybrid, or unknown.
Notice and reason: Meaningful notice should distinguish policy demotion from ordinary ranking variation.
Appeal and evidence: Provide status and reason without revealing exploitable security detail.
Measured impact: Use impressions, recommendation eligibility, rank distributions, and counterfactual limits—not anecdotes alone.
Alternative explanations: Interest change, competition, posting time, quality, network effects, or measurement error.
Documented safety purpose: Quality, relevance, misinformation risk, or abuse reduction.

CLVIS-007-REDUCED-DISTRIBUTION

Reduced distribution

A broad outcome category for reduced impressions or delivery that must be decomposed into mechanism and measurement.

Accountability record

Responsible actor: Platform, intermediary, user settings, or ordinary audience dynamics.
Written rule: Do not assume a policy action without a documented rule or status signal.
Decision mode: May result from many systems, manual action, or no intervention.
Notice and reason: Notice should identify a deliberate intervention if one occurred.
Appeal and evidence: Traffic data, eligibility, policy notices, and controlled comparisons may be relevant.
Measured impact: Traffic decline alone is not proof of suppression.
Alternative explanations: Audience change, seasonality, content mix, external events, competition, or analytics changes.
Documented safety purpose: Varies; document rather than infer.

CLVIS-008-DEMONETIZATION

Demonetization or monetization change

Advertising, subscription, tipping, recommendation, or revenue eligibility changes while content may remain hosted.

Accountability record

Responsible actor: Platform, advertiser, payment provider, or rights holder.
Written rule: Record monetization standard and affected product.
Decision mode: Automated, human, hybrid, advertiser-driven, or unknown.
Notice and reason: State the rule, affected revenue stream, duration, and appeal.
Appeal and evidence: Revenue and eligibility records should be available to the affected account.
Measured impact: Do not infer income loss without account-authorized records.
Alternative explanations: Advertiser demand, market rates, copyright claims, product changes, or audience shift.
Documented safety purpose: Brand safety, fraud prevention, policy compliance, or rights management.

CLVIS-009-LABELING

Labeling or contextualization

A warning, fact-check, provenance, sensitivity, or context label is attached to content.

Accountability record

Responsible actor: Platform, fact-checker, public authority, publisher, or provenance provider.
Written rule: Record label rule and evidence standard.
Decision mode: Automated, human, third-party, or hybrid.
Notice and reason: Explain why the label appears and whether it affects distribution.
Appeal and evidence: Link evidence and correction process.
Measured impact: A label's presence does not establish attention, belief, or behavior change.
Alternative explanations: Publisher metadata, user settings, legal notices, or accessibility context.
Documented safety purpose: Context, authenticity, graphic warning, or misinformation resilience.

CLVIS-010-SYNTHESIZED-ANSWERS

Synthesized summaries and AI answers

A service generates a summary or answer that may precede, replace, or frame source links.

Accountability record

Responsible actor: Search, assistant, platform, or publisher deploying the synthesis.
Written rule: Record source selection, attribution, correction, and refusal policy where available.
Decision mode: Generative model with retrieval, ranking, safety, and product layers.
Notice and reason: Identify synthetic output and material uncertainty.
Appeal and evidence: Provide citations, source inspection, feedback, and correction channels.
Measured impact: Presentation order is not proof of persuasion or displacement without exposure and behavior evidence.
Alternative explanations: Source availability, query wording, freshness, regional index, or ordinary generation error.
Documented safety purpose: Convenience, relevance, safety, and information synthesis.

CLVIS-011-PERSONALIZATION

Personalization

Different users receive different ordering, selection, or responses based on context or profile.

Accountability record

Responsible actor: Platform/service and profile/data controllers.
Written rule: Record personalization purpose, signals, and opt-out where applicable.
Decision mode: Algorithmic with possible human-curated constraints.
Notice and reason: Users should know material personalization is occurring and have meaningful controls.
Appeal and evidence: Profile access, signal correction, and non-personalized alternatives support contestability.
Measured impact: Different output does not itself establish manipulation or effect.
Alternative explanations: Language, location, device, session context, A/B testing, or randomization.
Documented safety purpose: Relevance, accessibility, safety, and product utility.

CLVIS-012-ACCOUNT-PENALTY

Account-level or content-level penalty

A strike, reduced functionality, posting limit, suspension, or reputation penalty applies to content or account.

Accountability record

Responsible actor: Platform/service.
Written rule: Record penalty rule, escalation ladder, and expiration.
Decision mode: Automated, human, hybrid, or undisclosed.
Notice and reason: State the triggering content, rule, penalty, duration, and appeal path.
Appeal and evidence: Preserve access to the challenged item and account history.
Measured impact: Penalty may affect future reach; document rather than assume.
Alternative explanations: Security lock, compromised account, rate limit, payment, or user setting.
Documented safety purpose: Repeat-abuse deterrence, security, or integrity.

CLVIS-013-AUTOMATED-REFUSAL

Automated refusal or response shaping

An AI system declines, redirects, narrows, or reframes a requested response.

Accountability record

Responsible actor: AI provider/deployer and policy owner.
Written rule: Record safety, legal, privacy, or product rule where disclosure is safe.
Decision mode: Model, classifier, policy layer, human escalation, or hybrid.
Notice and reason: A refusal should identify itself as a system boundary rather than impersonate the user's own belief.
Appeal and evidence: Provide feedback and correction for false or inconsistent refusals where appropriate.
Measured impact: A refusal does not establish belief change or suppression outside the interaction.
Alternative explanations: Capability limit, ambiguous request, outage, context-window limit, or policy uncertainty.
Documented safety purpose: Prevent harm, illegality, privacy invasion, or unsafe professional substitution.

CLVIS-014-MEMORY-PROFILE

Saved-memory update, deletion, or profile change

A system stores, edits, infers, deletes, or uses a persistent profile or memory about a person.

Accountability record

Responsible actor: AI provider/deployer and data controller.
Written rule: Record retention, training, personalization, deletion, and sharing policy.
Decision mode: User-directed, automatic, inferred, human-edited, or undisclosed.
Notice and reason: Notify users of material profile changes and give inspect/edit/delete controls.
Appeal and evidence: Provide accessible profile history and correction paths while protecting others' privacy.
Measured impact: A saved attribute does not prove identity, intention, or future behavior.
Alternative explanations: Session context, transient cache, user instruction, model inference, or import from another service.
Documented safety purpose: Continuity, personalization, security, and user convenience.

REMEDY AS EVIDENCE

A formal appeal is not meaningful unless it can repair the consequence

Contestability is tested through discoverability, specificity, authority, timeliness, accessibility, language access, restoration, downstream propagation, non-retaliation, and repeated-error prevention.

CLREM-001-NOTICE

Timely, understandable notice

Effective when: Delivered before or promptly after a consequential action, in plain language, through an accessible channel the person can retain.

Weak or failed when: Hidden in generic terms, delivered after the appeal deadline, or omits the action and responsible institution.

Evidence to retain: Timestamp, channel, language, accessibility, action, scope, duration, and contact.
CLREM-002-DATA-AND-RULE-ACCESS

Access to relevant data and rule

Effective when: The person can inspect the source data, inferred data, rule version, and evidence used, subject to bounded privacy/security redactions.

Weak or failed when: Only a generic category or unexplained score is provided.

Evidence to retain: Data fields, provenance, rule text, model/deployer role, redactions, and request outcome.
CLREM-003-SPECIFIC-EXPLANATION

Specific explanation

Effective when: Explains the principal reasons, rule, evidence, uncertainty, and role of automation sufficiently to challenge the outcome.

Weak or failed when: Model complexity, trade secrecy, or a boilerplate code substitutes for an actual reason.

Evidence to retain: Reason specificity, consistency with record, automation role, and understandable alternatives.
CLREM-004-CORRECTION

Correction of source and inferred data

Effective when: Both inaccurate inputs and unsupported inferences can be corrected, annotated, or suppressed, with provenance preserved.

Weak or failed when: Only the visible profile changes while downstream copies or decision records remain untouched.

Evidence to retain: Original value, correction, authority, downstream recipients, propagation confirmation, and exceptions.
CLREM-005-DELETION-RETENTION

Deletion and retention control

Effective when: Retention periods, legal exceptions, backups, model-training use, and deletion propagation are disclosed and enforceable.

Weak or failed when: A front-end deletion leaves operational profiles, biometric templates, or downstream datasets active.

Evidence to retain: Deletion request, systems covered, completion date, residual legal basis, and verification.
CLREM-006-HUMAN-AUTHORITY

Authorized human review

Effective when: A trained reviewer can inspect evidence, consider context/accommodations, overturn the result, and repair downstream effects.

Weak or failed when: The reviewer repeats the model score, lacks authority, or the appeal is closed automatically.

Evidence to retain: Reviewer role, authority, evidence inspected, decision, rationale, and conflict controls.
CLREM-007-INDEPENDENT-APPEAL

Independent and discoverable appeal

Effective when: The channel is easy to find, accessible, free or proportionate, and reviewed independently from the initial decision path.

Weak or failed when: The appeal repeats the same classifier, is unavailable in the person's language, or cannot change the outcome.

Evidence to retain: Discovery path, completion rate, reviewer independence, reversal rate, and reasons—not reversal rate alone.
CLREM-008-TIMELINESS

Response time and interim protection

Effective when: Urgency, livelihood, education, liberty, safety, and election/crisis context shape deadlines and interim relief.

Weak or failed when: A successful appeal arrives after the event, job, exam, benefit, or audience opportunity has passed.

Evidence to retain: Submission, acknowledgment, review, decision, restoration, and propagation timestamps.
CLREM-009-RESTORATION-REPAIR

Restoration, compensation, and downstream repair

Effective when: The remedy restores access or opportunity, removes erroneous strikes/labels, corrects downstream records, and addresses measurable loss where authorized.

Weak or failed when: Content returns but recommendation eligibility, reputation, pay, grade, or third-party records remain impaired.

Evidence to retain: Restored state, downstream systems, monetary/equitable relief, and residual harm.
CLREM-010-AUDIT-REPEAT-PREVENTION

Audit logs and repeated-error prevention

Effective when: Systems preserve accountable logs, investigate root causes, update rules/models/training, and test whether the error recurs across languages and groups.

Weak or failed when: A single case is fixed without identifying systemic causes or affected peers.

Evidence to retain: Version, trigger, reviewer path, root cause, corrective action, regression test, and aggregate outcome.
CLREM-011-ACCESSIBILITY-LANGUAGE

Accessibility, language, and advocate support

Effective when: Notice and remedy work with assistive technology, narrow screens, plain language, relevant languages, and authorized representatives.

Weak or failed when: The formal channel is unusable because of disability, literacy, language, identity verification, cost, or device barriers.

Evidence to retain: Languages, formats, assistive-technology tests, representative support, and failure/abandonment data.
CLREM-012-NONRETALIATION

Non-retaliation and safe correction

Effective when: People can question data and decisions without losing work, service, grades, care, benefits, or safety.

Weak or failed when: Appeal itself becomes a negative signal or requires disclosure that creates new risk.

Evidence to retain: Retaliation protections, complaint confidentiality, adverse changes after appeal, and independent oversight.
CLREM-013-TRANSPARENCY

Aggregate public transparency

Effective when: Aggregate reports disclose action types, reasons, automation, appeals, reversals, timing, language/region, and limitations without exposing individuals.

Weak or failed when: A single total hides mechanisms, groups, or whether users could obtain remedy.

Evidence to retain: Denominators, definitions, coverage, missingness, subgroup privacy, and changes over time.
CLREM-014-WITHDRAWAL-EXIT

Withdrawal, exit, and relationship continuity

Effective when: Users can pause, export, delete, disengage, or transfer without coercive friction, manipulative guilt, or silent loss of critical data.

Weak or failed when: Leaving triggers emotional pressure, irreversible profile loss, or continued use of private data beyond disclosed retention.

Evidence to retain: Exit path, data export, memory deletion, subscription effects, crisis routing, and post-exit retention.

OUTCOME & DOWNSTREAM REPAIR

What changed after a challenge—and what remained unresolved

These case files separate a remedy requirement from its observed result. Restoration, monetary relief, deletion, policy correction, external review, and legal implementation are recorded independently from lost reach, copied signals, reputation effects, delayed access, and repeated-error prevention.

CLOUT-001-RITE-AID-DOWNSTREAM-DELETIONRite Aid facial-recognition order: use ban, deletion, and third-party propagation REGULATORY_ORDER_WITH_DOWNSTREAM_REPAIR_REQUIREMENTS

A modified order imposed a five-year facial-recognition surveillance-use ban and specified deletion, monitoring, notice, complaint-response, and third-party propagation duties.

Responsible institution: U.S. Federal Trade Commission and the parties bound by the modified order
Notice and discoverability: The public complaint, case page, and order are discoverable; individual consumer notice duties are bounded by the order.
Rule-specific reason: The FTC alleged unfair deployment and inadequate safeguards under the FTC Act and order violations.
Correction result: The order includes complaint-response and review duties; no aggregate individualized correction success rate is recorded here.
Deletion or retention result: Covered information and derived models/algorithms were ordered deleted or destroyed, subject to legal limits and sworn reporting.
Human authority and independence: The order requires trained personnel and monitoring, but this record does not claim every individual review was independent or outcome-changing. Regulatory oversight is external to the retailer; no universal consumer appeal tribunal is created.
Repair: Use prohibition, deletion, assessment, notice, complaint handling, and monitoring are concrete remedies; compensation is not recorded as a universal remedy.
Downstream propagation: The order requires identification of third parties, deletion instructions, demands for written confirmation, and reporting of confirmations or responses.
Repeat-error prevention: A mandated biometric monitoring program, testing, safeguards, and reporting seek to prevent recurrence.
Unresolved downstream harm: Third-party deletion completion, reputation effects, distress, lost access, and every copied or derived artifact are not independently verified here.
Evidence boundary: A settlement/order is not a contested-trial finding; a required remedy is not proof of completed downstream repair.
Reopening trigger: Reopen on public compliance reports, enforcement modification, verified third-party deletion results, or subsequent litigation.

CLOUT-002-ITUTORGROUP-EMPLOYMENT-RELIEFiTutorGroup automated age-screening settlement EMPLOYMENT_DISCRIMINATION_SETTLEMENT_WITH_MONETARY_AND_INJUNCTIVE_RELIEF

The parties resolved an EEOC suit alleging software automatically rejected older applicants; the settlement provided $365,000 for more than 200 applicants and multi-year non-monetary relief.

Responsible institution: U.S. Equal Employment Opportunity Commission, federal court, and settling employers
Notice and discoverability: The public EEOC announcement explains the allegation and settlement; applicant-level notice quality before litigation is not established.
Rule-specific reason: The EEOC alleged age and sex discrimination under the ADEA based on programmed age thresholds.
Correction result: The settlement repairs a defined class through monetary relief and prospective controls; individual application reconsideration is not claimed for every person.
Deletion or retention result: No comprehensive applicant-data deletion outcome is recorded in the reviewed source.
Human authority and independence: Prospective oversight and training relief were imposed; this record does not claim a human reviewer corrected every past decision. EEOC litigation and court supervision provided external review; ordinary applicant appeals remain deployment-specific.
Repair: $365,000 and non-monetary relief are documented; lost work, delay, and every downstream consequence were not quantified.
Downstream propagation: No verified correction propagation to every recruiting database, vendor, or future model is recorded.
Repeat-error prevention: Five-year decree and non-monetary relief were reported; exact compliance outcomes remain separately reviewable.
Unresolved downstream harm: Lost earnings, employment history, confidence, and copied screening records may persist beyond monetary relief.
Evidence boundary: Settlement allegations and relief are documented; the case did not produce a contested merits judgment on every fact.
Reopening trigger: Reopen on decree compliance reports, later court orders, or verified applicant repair evidence.

CLOUT-003-DSA-OUT-OF-COURT-AGGREGATEDSA out-of-court dispute outcomes in the first half of 2025 AGGREGATE_PROCEDURAL_REMEDY_OUTCOME

The Commission reports more than 1,800 disputes reviewed in the first half of 2025 and platform decisions reversed in 52% of closed cases, restoring content or accounts.

Responsible institution: Certified EU out-of-court dispute settlement bodies, platforms, national coordinators, and European Commission reporting
Notice and discoverability: DSA processes require reasons and redress routes, but accessibility and discoverability vary by service and person.
Rule-specific reason: Case-specific reasons are handled within disputes; the aggregate source does not publish every reason or evidence file.
Correction result: Closed cases had a reported 52% reversal rate in the cited aggregate; individual categories and denominator details remain source-bounded.
Deletion or retention result: The aggregate does not establish deletion of every moderation profile, strike, or copied ranking signal.
Human authority and independence: Certified bodies provide external process; authority and enforceability differ by mechanism and case. Out-of-court bodies are structurally external to platforms, but this record does not evaluate each body’s practical independence.
Repair: Restoration of content and accounts is reported; compensation, reach recovery, and reputational repair are not established.
Downstream propagation: No aggregate proof shows that all strikes, recommender signals, mirrors, search caches, or monetization records were corrected.
Repeat-error prevention: Aggregate reversals can inform oversight, but systemic prevention must be evidenced separately.
Unresolved downstream harm: Lost time, reach, income, audience trust, and copied enforcement signals may outlast restoration.
Evidence boundary: A successful appeal in one or many submitted cases cannot be generalized to all moderation decisions.
Reopening trigger: Reopen when certified-body datasets publish category, service, language, timeliness, accessibility, and downstream-repair details.

CLOUT-004-BREAST-CANCER-RESTORATIONBreast-cancer-awareness content: fifteen acknowledged enforcement errors CASE_BUNDLE_RESTORATION_AFTER_EXTERNAL_ESCALATION

Meta restored all fifteen breast-cancer-awareness posts after the Board brought the appeals to the company.

Responsible institution: Meta and the Oversight Board
Notice and discoverability: Users reached the Board appeal process; the source does not establish that every affected creator could discover or access it.
Rule-specific reason: The bundle concerns mistaken enforcement against medical-awareness imagery under nudity-related rules.
Correction result: All fifteen appealed posts were restored.
Deletion or retention result: No evidence establishes deletion of strikes, model features, or derived enforcement signals beyond the documented correction.
Human authority and independence: External escalation changed the outcome; ordinary first-line reviewer authority is not demonstrated. The Board is structurally separate from Meta but depends on the platform’s case framework and implementation.
Repair: Post restoration is confirmed; reach, campaign timing, health-information access, and monetization repair are not measured.
Downstream propagation: No public proof shows correction of every recommender, strike, cache, or duplicate signal.
Repeat-error prevention: The decision identifies improvement opportunities but does not establish completed systemic prevention.
Unresolved downstream harm: Time-sensitive awareness reach and audience trust may not be recoverable after restoration.
Evidence boundary: Restoration confirms correction of these decisions, not the full causal chain of lost reach or health outcomes.
Reopening trigger: Reopen on Meta implementation evidence, repeat-error data, or creator-reported downstream repair outcomes.

CLOUT-005-SOMALILAND-JOURNALISM-RESTORATIONSomaliland journalism page, four posts, and strike restored MULTI_LAYER_ACCOUNT_CONTENT_AND_STRIKE_RESTORATION

Meta republished a Somali-language journalism page, restored four posts, reversed the account strike, and later reinstated additional Somaliland content it acknowledged was removed in error.

Responsible institution: Meta and the Oversight Board
Notice and discoverability: Four post appeals received repeated human review; the page appeal was automatically closed without prioritized review before Board escalation.
Rule-specific reason: The page and posts were incorrectly treated as violating Hateful Conduct despite public-interest journalism context.
Correction result: Page, posts, and strike were restored; ten additional Somaliland appeals were also reported as errors and reinstated.
Deletion or retention result: No public record confirms deletion of all prior risk labels or copied moderation signals.
Human authority and independence: Six human reviews upheld errors; external Board escalation prompted reversal, showing that human review alone did not guarantee remedy. The Board supplied external review after internal and automated appeal paths failed.
Repair: Content, page, and strike restoration are concrete; lost audience contact, news timeliness, revenue, and journalist safety effects remain unmeasured.
Downstream propagation: No complete propagation record covers search visibility, follower feeds, recommendations, mirrors, or future review queues.
Repeat-error prevention: The decision recommends stronger journalism and language safeguards; implementation remains separately evidentiary.
Unresolved downstream harm: News timeliness, safety, source trust, audience reach, and future account risk may persist after reinstatement.
Evidence boundary: The case establishes documented error and restoration, not full downstream or population-level effect.
Reopening trigger: Reopen on Meta implementation updates, repeated Somali-language error data, or creator/press-freedom outcome evidence.

CLOUT-006-KENYA-SLUR-CURRENTNESSKenyan political speech restored after slur-list currentness review POLICY_CLASSIFICATION_CORRECTION

The Board overturned removal of a Kenyan political comment and found the contested term should not have qualified as a slur at the time of posting.

Responsible institution: Meta and the Oversight Board
Notice and discoverability: The user reached external appeal; ordinary users without escalation may face different notice and access conditions.
Rule-specific reason: Hateful Conduct / slur designation was applied too broadly to evolving political language.
Correction result: The removal decision was overturned and content restored.
Deletion or retention result: No proof establishes deletion of all policy-risk labels or downstream ranking effects.
Human authority and independence: External review changed the result; reviewer authority and policy-list governance remain distinct. The Board provides independent judgment but is not a court or universal public regulator.
Repair: Content restoration and policy correction are documented; political reach and debate timing were not restored measurably.
Downstream propagation: No published evidence confirms every language list, classifier, strike, recommender, or reviewer tool was updated.
Repeat-error prevention: The Board urges judicious, current designation practice; implementation remains open.
Unresolved downstream harm: Lost election-period attention, account trust, and self-censorship may persist.
Evidence boundary: Contextual reversal does not mean the term is never harmful or never regulable.
Reopening trigger: Reopen on public evidence that Meta updated the designation process and tested downstream language effects.

CLOUT-007-SERCO-BIOMETRIC-DELETIONSerco employee-attendance biometrics: stop-processing and destruction order DATA_PROTECTION_ENFORCEMENT_WITH_CESSATION_AND_DELETION

The ICO ordered covered entities to stop biometric attendance processing and destroy biometric data not legally required within the specified compliance period.

Responsible institution: UK Information Commissioner's Office and covered Serco Leisure entities
Notice and discoverability: The public enforcement notice is discoverable; employee-level notice and remedy accessibility are not fully recorded.
Rule-specific reason: The ICO found the processing was neither necessary nor proportionate and workers lacked a proactively offered alternative.
Correction result: The remedy is cessation and destruction rather than correction of an inference; individual complaints are not enumerated.
Deletion or retention result: Destruction of data not legally required was ordered within three months.
Human authority and independence: Regulatory authority can compel cessation; internal workplace review authority was insufficient to prevent the practice. The ICO is external to the employer; legal appeal routes remain separate.
Repair: Stop-processing and destruction are concrete; compensation, workplace trust, and every copied record are not established.
Downstream propagation: The public source does not fully map every processor, vendor, backup, or derivative template.
Repeat-error prevention: Necessity, proportionality, alternatives, and data-protection impact assessment requirements are emphasized.
Unresolved downstream harm: Copied templates, attendance histories, workplace consequences, and trust effects may persist without separate evidence.
Evidence boundary: An enforcement order requiring destruction is not proof every copy was destroyed or every worker repaired.
Reopening trigger: Reopen on compliance confirmation, appeal, employee evidence, or processor-level deletion documentation.

CLOUT-008-INTELLIVISION-VALIDATION-ORDERIntelliVision consent order: substantiation and demographic-performance claims MARKETING_AND_VALIDATION_GOVERNANCE_ORDER

The FTC order restricts unsubstantiated claims about facial-recognition accuracy, demographic performance, and liveness/spoofing and requires competent, reliable, documented testing.

Responsible institution: U.S. Federal Trade Commission and IntelliVision Technologies Corp.
Notice and discoverability: The consent order is public; consumer-level pre-purchase understanding is not measured.
Rule-specific reason: The FTC alleged misleading or unsupported accuracy, no-bias, training-data, and anti-spoofing representations.
Correction result: The remedy governs future claims; individual false-match corrections are not the focus.
Deletion or retention result: No general deletion remedy is documented in the reviewed order.
Human authority and independence: Qualified testing professionals and documented review are required for claims; no universal affected-person appeal process is created. FTC oversight is external; product users do not receive an independent merits tribunal through this order alone.
Repair: Marketing substantiation and documentation are corrected; no proof of restored opportunities or compensated users is recorded.
Downstream propagation: No complete record shows correction of every reseller, integration, archived claim, or deployment decision.
Repeat-error prevention: Objective qualified testing and documentation requirements seek to prevent repeat unsupported claims.
Unresolved downstream harm: Past purchasing and deployment decisions based on unsupported claims may persist.
Evidence boundary: Substantiated marketing claims are not equivalent to construct validity, deployment safety, or field effectiveness.
Reopening trigger: Reopen on public compliance testing, independent field evaluations, or later enforcement.

CLOUT-009-CHILE-EMOTIV-IMPLEMENTATION-GAPChile Emotiv Insight case: judicial order and partial regulatory implementation APPELLATE_RIGHTS_RULING_WITH_IMPLEMENTATION_GAP

The Supreme Court required public-authority evaluation and compliant handling of brain data; later ISP review concluded the consumer device was outside its then-current medical-device competence.

Responsible institution: Chile Supreme Court, Instituto de Salud Pública, customs authority, and Emotiv as described in the public record
Notice and discoverability: The ruling and parliamentary research are public; ordinary consumer notice and product transparency remain separate questions.
Rule-specific reason: The case concerned constitutional rights, novel neurotechnology, commercial use, and brain-data handling.
Correction result: The reported judicial remedy included compliant data handling and deletion in the individual case; independent completion evidence is incomplete.
Deletion or retention result: Deletion was ordered/reported at case level; downstream copies and derived models are not comprehensively verified.
Human authority and independence: Judicial review changed the legal posture; the ISP’s competence boundary limited regulatory follow-through. Supreme Court review was independent of the company; implementation depended on public authorities with bounded statutory competence.
Repair: Judicial recognition and evaluation/deletion directions are concrete; comprehensive product-market governance and downstream data repair remain incomplete.
Downstream propagation: No complete record verifies deletion from all vendor systems, processors, backups, analytics, or models.
Repeat-error prevention: The case prompted review and illustrates need for anticipatory authority, but the competence gap remains.
Unresolved downstream harm: Data-copy status, product changes, consumer understanding, and future oversight remain unresolved.
Evidence boundary: A landmark individual ruling is not a complete implementation regime or universal neurorights code.
Reopening trigger: Reopen on compliance proof, new Chilean implementing law/regulation, ISP competence changes, or further judgments.

CLOUT-010-SAFERENT-HOUSING-SETTLEMENTSafeRent tenant-screening settlement: compensation and score restrictions COURT_APPROVED_CLASS_SETTLEMENT_WITH_PRODUCT_RESTRICTIONS

A court-approved settlement provided $2.275 million and product restrictions for a defined class of Massachusetts housing-voucher applicants; payments were distributed in 2025.

Responsible institution: U.S. District Court, settlement administrator, SafeRent Solutions, and class representatives
Notice and discoverability: The settlement site published deadlines and distribution status; notice at the original housing decision was a disputed harm rather than a proven adequate remedy.
Rule-specific reason: Plaintiffs alleged tenant-screening scores disproportionately harmed voucher users and protected groups; SafeRent denied wrongdoing and settled.
Correction result: Eligible class members could claim payment; product restrictions address future scoring, not complete reconsideration of every past housing decision.
Deletion or retention result: No comprehensive deletion of applicant data, landlord copies, credit files, or derived profiles is established.
Human authority and independence: Future individualized landlord assessment is encouraged by score restrictions, but human review quality and authority are not universally guaranteed. Court supervision and class counsel supplied independent process; ordinary tenant-screening disputes remain sector-specific.
Repair: Payments and product restrictions are concrete; lost housing, displacement, search costs, credit effects, and emotional harm are not fully repaired.
Downstream propagation: No proof confirms correction across every landlord, property manager, tenant-screening report, credit record, or downstream decision.
Repeat-error prevention: Five-year restrictions and validation conditions seek to reduce recurrence for covered voucher applicants.
Unresolved downstream harm: Housing instability, copied reports, landlord beliefs, and opportunity loss may persist.
Evidence boundary: Settlement is not an admission or merits judgment; relief is bounded to the agreement and class.
Reopening trigger: Reopen on compliance reports, independent validation, subsequent litigation, or class-member outcome evidence.

CLOUT-011-CFPB-GUIDANCE-WITHDRAWALCFPB complex-algorithm adverse-action circular withdrawn while underlying duties remain LEGAL_GUIDANCE_SUPERSESSION_CURRENTNESS

Circular 2022-03 and related 2023 guidance were withdrawn on 2025-05-12; ECOA and Regulation B remain separate statutory/regulatory authorities.

Responsible institution: Consumer Financial Protection Bureau
Notice and discoverability: The CFPB publishes a withdrawn-guidance index and archives prior circular content.
Rule-specific reason: Agency guidance status changed; the underlying adverse-action statutory framework was not repealed by the withdrawal record.
Correction result: WIP.54 corrects the public legal marker and source currentness; no person-level adverse-action remedy is claimed.
Deletion or retention result: NOT_APPLICABLE_TO_GUIDANCE_STATUS
Human authority and independence: Covered creditors remain responsible under applicable law; agency interpretive posture and judicial review are separate. Legal complaint and court routes remain separate from the circular’s status.
Repair: The repository corrects a legal-status description; it does not repair any consumer credit decision.
Downstream propagation: All site claims, legal markers, source records, and durable memory must carry the withdrawal status.
Repeat-error prevention: Currentness fields and reopening triggers prevent archived guidance from being silently presented as current.
Unresolved downstream harm: Third-party publications may continue citing withdrawn circulars without status labels.
Evidence boundary: Withdrawal of guidance is not repeal of underlying statute or regulation and does not decide every creditor obligation.
Reopening trigger: Reopen on replacement guidance, rulemaking, enforcement, or controlling court interpretation.

CLOUT-012-EU-AI-ACT-IMPLEMENTATION-2026EU AI Act implementation: enacted 2026 Omnibus and Article 50 application ENACTED_LEGAL_IMPLEMENTATION_CURRENTNESS

Regulation (EU) 2026/1744 entered into force on 2026-07-27; Article 50 transparency duties apply from 2026-08-02, while selected high-risk implementation dates were extended.

Responsible institution: European Union legislature, European Commission, AI Office, and national competent authorities
Notice and discoverability: Official Journal, Commission timeline, guidance, and code materials are public; deployer-level notice depends on article, role, and use.
Rule-specific reason: The implementation framework addresses risk categories, prohibited practices, transparency, general-purpose AI, and high-risk systems under defined scope and exceptions.
Correction result: WIP.54 corrects the prior status from proposed to enacted amendments and updates application dates.
Deletion or retention result: No universal deletion remedy follows solely from this implementation marker.
Human authority and independence: Human oversight requirements apply to defined systems; actual authority and effectiveness require deployment evidence. Remedies derive from the AI Act, sector law, data protection, consumer law, labor law, and national procedure rather than one universal appeal body.
Repair: Currentness correction improves legal accuracy; no affected-person repair outcome is claimed.
Downstream propagation: All dependent site claims and memory records must distinguish enacted amendments, current application dates, and delayed high-risk provisions.
Repeat-error prevention: Date-specific legal fields and reopening triggers reduce proposal/enactment confusion.
Unresolved downstream harm: Compliance quality, enforcement consistency, and remedy outcomes remain unmeasured across deployments.
Evidence boundary: Enactment and application dates do not establish compliance, effectiveness, or universal legal coverage.
Reopening trigger: Reopen on delegated acts, standards, enforcement decisions, court rulings, or further amendment.

JURISDICTION & CURRENTNESS

Enacted text, effective dates, guidance, inquiries, and rulings remain separate

The records below were reviewed for WIP.54. They are educational summaries, not legal advice, and do not convert one jurisdiction’s rule into a universal cognitive-liberty code.

CLLAW-001-INTERNATIONAL-FORUM-INTERNUMInternational human-rights law · Freedom of thought and the forum internumESTABLISHED_FOUNDATION_WITH_DEVELOPING_AI_APPLICATION

Enacted text: ICCPR Article 18 protects freedom of thought, conscience, and religion; treaty obligations attach according to ratification and international-law rules.
Effective date: ICCPR entered into force 1976-03-23; state-specific status varies.
Implementing regulation: Not a single global implementing regulation.
Official guidance: UN Special Rapporteur A/76/380 analyzes the forum internum and technological threats.
Complaint or allegation: NOT_APPLICABLE_TO_THIS_GENERAL_MARKER
Enforcement or ruling: No single universal AI-inference judgment recorded here.
Scope limit: International interpretation is not self-executing domestic advice and does not create one uniform cognitive-liberty cause of action.
Reviewed: 2026-07-28

CLLAW-002-UNESCO-NEUROTECHUNESCO member-state normative framework · Ethics of neurotechnologyADOPTED_NORMATIVE_RECOMMENDATION_NOT_BINDING_TREATY

Enacted text: UNESCO General Conference adopted a Recommendation on the Ethics of Neurotechnology in November 2025.
Effective date: Official UNESCO materials state entry into force on 2025-11-12 within UNESCO's recommendation framework.
Implementing regulation: Member-state implementation is separate and jurisdiction-specific.
Official guidance: The text addresses dignity, autonomy, privacy, consent, accountability, equity, and misuse.
Complaint or allegation: NOT_APPLICABLE_TO_NORMATIVE_INSTRUMENT
Enforcement or ruling: No universal judicial enforcement mechanism created by the recommendation.
Scope limit: Do not label it enacted domestic law or proof of implementation.
Reviewed: 2026-07-28

CLLAW-003-CHILEChile · Brain activity and information derived from itENACTED_CONSTITUTIONAL_REFORM_WITH_CASE_LEVEL_RULING_AND_PARTIAL_IMPLEMENTATION

Enacted text: Law 21.383 amended constitutional language concerning scientific and technological development and special protection of brain activity and information derived from it.
Effective date: Published 2021-10-25.
Implementing regulation: The constitutional reform remains in force. Sector implementation is incomplete; reviewed BCN material records a Supreme Court evaluation order and an ISP conclusion that Emotiv Insight was outside its then-current medical-device competence.
Official guidance: Official BCN 2026 research distinguishes the constitutional protection, pending/sector legislation, the Supreme Court ruling, and the ISP competence finding.
Complaint or allegation: Girardi v. Emotiv alleged rights violations and unconsented brain-data retention; allegation, ruling, and implementation are recorded separately.
Enforcement or ruling: No universal interpretation or complete enforcement map asserted here.
Scope limit: The case does not create a complete consumer-neurotechnology regulator, verify all deletion, or govern ordinary non-neural behavioral inference or other countries.
Reviewed: 2026-07-28

CLLAW-004-COLORADOColorado, United States · Biological and neural data under state privacy lawENACTED_STATE_PRIVACY_PROTECTION

Enacted text: HB24-1058 expanded Colorado privacy protections for biological data, including neural data within statutory definitions.
Effective date: 2024-08-07.
Implementing regulation: Colorado Privacy Act rules and enforcement structure apply according to their scope.
Official guidance: Official bill text and legislative status are the source authority.
Complaint or allegation: No enforcement complaint relied upon here.
Enforcement or ruling: No universal case-law conclusion recorded.
Scope limit: Entity, data, exemption, enforcement, and consumer-right scope must be checked before application.
Reviewed: 2026-07-28

CLLAW-005-CALIFORNIACalifornia, United States · Neural data as sensitive personal informationENACTED_STATE_CONSUMER_PRIVACY_PROTECTION

Enacted text: SB 1223 added neural data to sensitive personal information under California consumer-privacy law.
Effective date: 2025-01-01.
Implementing regulation: California Privacy Protection Agency regulations and statutory exemptions remain relevant.
Official guidance: Official chaptered bill text is the source authority.
Complaint or allegation: No case-specific complaint relied upon here.
Enforcement or ruling: No complete enforcement or appellate map asserted.
Scope limit: Does not cover every entity, inference, or use and is not a universal mental-privacy code.
Reviewed: 2026-07-28

CLLAW-006-EU-AI-ACTEuropean Union · AI Act manipulation, emotion recognition, transparency, employment, education, and law-enforcement scopeENACTED_REGULATION_WITH_ENACTED_2026_AMENDMENTS_AND_PHASED_APPLICATION

Enacted text: Regulation (EU) 2024/1689 is enacted and directly applicable according to its phased timetable, definitions, exceptions, and sector-specific provisions.
Effective date: AI Act entered into force 2024-08-01. Prohibited practices applied 2025-02-02; GPAI/governance rules applied 2025-08-02; Article 50 transparency duties apply 2026-08-02; selected high-risk dates were extended by Regulation (EU) 2026/1744.
Implementing regulation: Regulation (EU) 2026/1744 entered into force 2026-07-27. Commission guidance, codes, standards, authorities, and delegated/implementing acts remain part of staged implementation.
Official guidance: Commission timeline and Article 50 transparency materials reviewed through 2026-07-28.
Complaint or allegation: Not a case-specific marker.
Enforcement or ruling: No claim of final interpretation for every article or exception.
Scope limit: Do not describe the 2026 Omnibus as merely proposed. Article 50 application does not mean every high-risk obligation is already in force; exceptions and role-specific duties remain material.
Reviewed: 2026-07-28

CLLAW-007-EU-DSAEuropean Union · Statement of reasons, complaint, out-of-court dispute, recommender transparency, and systemic riskENACTED_AND_OPERATIONAL_WITH_AGGREGATE_IMPLEMENTATION_EVIDENCE

Enacted text: Regulation (EU) 2022/2065 establishes bounded procedural and systemic obligations for covered intermediary services.
Effective date: General application from 2024-02-17, with earlier designated-service obligations.
Implementing regulation: Delegated acts, Commission enforcement, national Digital Services Coordinators, and transparency databases support implementation.
Official guidance: Commission implementation materials report more than 1,800 first-half 2025 out-of-court disputes and 52% reversal among closed cases; this is not an all-decision error rate.
Complaint or allegation: Aggregate complaints are not equivalent to findings of platform error.
Enforcement or ruling: Commission investigations, fines, and court review must be recorded separately by case.
Scope limit: Not a universal global moderation code; a formal channel does not prove effective or accessible remedy for every user.
Reviewed: 2026-07-28

CLLAW-008-ILLINOIS-BIPAIllinois, United States · Biometric identifiers and biometric informationENACTED_WITH_2024_AMENDMENT_AND_ACTIVE_CASE_LAW_BOUNDARIES

Enacted text: BIPA imposes duties concerning specified biometric identifiers/information, notice, consent, retention/destruction, disclosure, and security; PA 103-0769 amended recovery and electronic-consent provisions.
Effective date: BIPA effective 2008; PA 103-0769 effective 2024-08-02.
Implementing regulation: No single comprehensive implementing regulation recorded here.
Official guidance: Current statute and public act are the primary authorities.
Complaint or allegation: Private complaints and class actions are allegations until resolved.
Enforcement or ruling: Case-specific trial and appellate holdings must be separately mapped.
Scope limit: Does not cover all behavioral, emotional, neural, or probabilistic inferences and is not legal advice.
Reviewed: 2026-07-28

CLLAW-009-ILLINOIS-EMPLOYMENT-AIIllinois, United States · AI in employment decisionsENACTED_STATUTE_EFFECTIVE; FINAL_DETAILED_RULEMAKING_NOT_RETRIEVED

Enacted text: Public Act 103-0804 amended the Illinois Human Rights Act to address discriminatory AI use, zip-code proxy use, and notice in employment decisions.
Effective date: 2026-01-01.
Implementing regulation: No official final detailed implementing rule was located in the bounded WIP.54 search. The retained secondary report says proposed rules were withdrawn; final official rule status is NOT_RETRIEVED.
Official guidance: Enacted text and official IDHR materials remain primary. Do not treat secondary withdrawal reporting as a final agency disposition.
Complaint or allegation: No case-specific complaint relied upon here.
Enforcement or ruling: No final appellate interpretation recorded.
Scope limit: Secondary reporting describes withdrawal of proposed rules; reopen on an official final, replacement, or refiled rulemaking record.
Reviewed: 2026-07-28

CLLAW-010-US-ADVERSE-ACTIONUnited States federal consumer-credit law · Specific reasons for algorithmic adverse actionGUIDANCE_WITHDRAWN; UNDERLYING_STATUTORY_AND_REGULATORY_DUTIES_REMAIN

Enacted text: ECOA and Regulation B adverse-action duties predate the circular; the circular states that complex algorithms do not excuse failure to provide specific principal reasons.
Effective date: Underlying statutory and regulatory duties apply according to their existing dates and scope.
Implementing regulation: Regulation B and official interpretations govern covered creditors.
Official guidance: CFPB Circular 2022-03 is withdrawn historical guidance as of 2025-05-12. Underlying ECOA and Regulation B duties remain separate legal authorities.
Complaint or allegation: No individual credit case relied upon here.
Enforcement or ruling: No universal judicial holding asserted.
Scope limit: Do not present Circular 2022-03 or Circular 2023-03 as current CFPB guidance. The withdrawal does not repeal ECOA/Regulation B or create a universal explanation right.
Reviewed: 2026-07-28

NORMATIVE PROPOSAL — NOT ENACTED LAW. JURISDICTION-SPECIFIC LAW — NOT A UNIVERSAL CODE. No named legal specialist has completed a disposition.

NORMATIVE PROPOSAL

Twelve declarations for a free mind

These are project proposals informed by the source record. They are not enacted law, a certification standard, or evidence that any institution has adopted them.

The inner life is sovereign.
Thought is not conduct.
Mental privacy is a human-rights issue.
A prediction is not a verdict.
An inference about emotion is not proof of emotion.
A historical record is not a permanent identity.
AI must identify itself and its role.
No system should engineer emotional dependency.
Refusal must be transparent rather than disguised as the user’s own belief.
Consequential automated decisions require explanation and appeal.
Safety must be specific, proportionate, and evidence-based.
Human responsibility cannot be delegated to a machine.

CONTEMPORARY / ONGOING CLAIM — NOT SETTLED HISTORY

SOURCE LINEAGE

Seven exact owner reports and a bounded external registry

Exact byte preservation establishes identity and lineage. It does not independently certify every embedded citation, causal claim, current legal conclusion, platform practice, or reproduction right.

Exact owner-supplied reports

The Mind Is Not a Battlespace: Cognitive Liberty in the Age of AI-Enabled Conflict Cognitive Liberty Policy Report.md · 49,764 bytes 38123286b4b1f926f4eb61d905da31996ac69cf1daeaeccf888fc9bffdf6c13a
Does not independently establish every embedded citation, current legal conclusion, causal effect, platform practice, or universal right.
The Mind Is the First Territory of Freedom The Mind Is the First Territory of Freedom.md · 31,597 bytes c0330b43e0e7760bc434ba432dcc08db981c0e9a54f6457e2f0a3854c71a9e8d
Does not independently establish every embedded citation, current legal conclusion, causal effect, platform practice, or universal right.
Keep the Human Mind Free AI Cognitive Liberty Manifesto.md · 39,620 bytes 0969e7800184e26b57fe0b2af9e2aba8c8ef46fda7f870d4bd04262cd5cc5962
Does not independently establish every embedded citation, current legal conclusion, causal effect, platform practice, or universal right.
The Invisible Editor: AI Censorship, Algorithmic Suppression, and the Right to Know The Invisible Editor AI Censorship, Algorithmic Suppression, and the Right to Know.md · 35,621 bytes 8de48a8e5a90d2792c789185d26e476308df7286b65979c04f38a091dbdce0ee
Does not independently establish every embedded citation, current legal conclusion, causal effect, platform practice, or universal right.
Cognitive Liberty Is the Civil-Rights Struggle of the AI Age Cognitive Liberty AI Speech Prep.md · 39,135 bytes e95edc2c83f609cf9d510eb76b822f2bfcfebeb2efe56df3c288f8542b55635a
Does not independently establish every embedded citation, current legal conclusion, causal effect, platform practice, or universal right.
Algorithmic Suppression and AI-Driven Censorship AI Moderation and Suppression Research.md · 56,439 bytes 46900cf830ecd5a36b4f574c23d133b898e91256177b55ca1626f1b9107c9430
Does not independently establish every embedded citation, current legal conclusion, causal effect, platform practice, or universal right.
When the System Watches the Mind AI Monitoring And Mental Privacy.md · 46,674 bytes 90646aacc27571eb852fcd668705bca53a52df350421136895794b27ff7b5d7f
Does not independently establish every embedded citation, current legal conclusion, causal effect, platform practice, or universal right.

Reviewed primary and authoritative records

A/76/380: Freedom of thought — report of the Special Rapporteur on freedom of religion or belief United Nations Office of the High Commissioner for Human Rights · Authoritative international human-rights interpretation
Does not itself enact a standalone global statutory right named cognitive liberty or resolve every application to AI inference.
International Covenant on Civil and Political Rights United Nations Office of the High Commissioner for Human Rights · Primary international legal instrument
Does not use the modern umbrella term cognitive liberty or specify all AI-era implementation duties.
Allied Command Transformation develops the Cognitive Warfare Concept NATO Allied Command Transformation · Official NATO command publication
Does not make every research paper formal NATO doctrine, prove universal national adoption, or authorize domestic population control.
Ley 21.383 — scientific and technological development in service of people Biblioteca del Congreso Nacional de Chile · Primary national legal text
Does not create a universal global cognitive-liberty code or settle all secondary implementing legislation.
HB24-1058 — Protect Privacy of Biological Data Colorado General Assembly · Primary state legislative record
Does not cover every mental-state inference, every entity, or a complete standalone cognitive-liberty right.
SB 1223 — Consumer privacy: sensitive personal information: neural data California Legislative Information · Primary state legislative record
Does not regulate all cognitive inference, workplace monitoring, or government use in one comprehensive code.
Regulation (EU) 2024/1689 — Artificial Intelligence Act European Union · Primary legal text
Does not prohibit all affective computing or apply identically outside EU scope and transition rules.
Regulation (EU) 2022/2065 — Digital Services Act European Union · Primary legal text
Does not eliminate moderation error, mandate identical platform ranking, or operate as a global speech code.
Illinois Public Act 103-0804 — AI in employment under the Human Rights Act Illinois General Assembly · Primary state legal text
Does not create a general right against every automated decision or every form of workplace monitoring.
Illinois Biometric Information Privacy Act, 740 ILCS 14 Illinois General Assembly · Primary state legal text
Does not cover all inferred emotions, thoughts, neural data, photographs, or every public-sector use.
Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements Psychological Science in the Public Interest · High-quality scientific synthesis
Does not imply all physiological or multimodal measurement is useless in every clinical or research context.
Artificial Intelligence Risk Management Framework (AI RMF 1.0) U.S. National Institute of Standards and Technology · Authoritative technical governance framework
Is voluntary and does not itself create legal rights or substitute for sector-specific law.
Meta Final Update: Israel and Palestine Human Rights Due Diligence Meta · First-party record describing independent due-diligence follow-up
Does not independently resolve all claims of bias, prove equal outcomes, or cover every Meta product and conflict context.
FTC Launches Inquiry into AI Chatbots Acting as Companions U.S. Federal Trade Commission · Primary regulator record
An inquiry is not an adjudication, final finding, or proof that every companion product causes dependency or harm.
Recommendation on the Ethics of Neurotechnology UNESCO · Authoritative UNESCO normative instrument; recommendation rather than binding treaty
Does not enact a binding universal cognitive-liberty statute, resolve domestic implementation, or establish that any particular inference system is accurate.
OECD Neurotechnology Toolkit Organisation for Economic Co-operation and Development · Authoritative OECD policy toolkit supporting a Council recommendation
Does not create binding law for all jurisdictions or prove that recommended safeguards have been implemented in any particular product or workplace.
AI Act regulatory framework and implementation timeline European Commission · Primary EU institutional currentness source
Does not make all obligations immediately applicable, eliminate exceptions, prove provider compliance, or provide legal advice for a particular deployment.
How the Digital Services Act enhances content moderation transparency and appeals European Commission · Primary EU institutional implementation source
Does not supply an all-decision denominator, platform-wide error rate, universal accessibility finding, or proof that every downstream strike, ranking, cache, income, or audience effect was repaired.
Illinois Department of Human Rights temporarily withdraws proposed AI employment rules Seyfarth Shaw LLP · Secondary legal analysis; underlying statute remains the primary authority
Does not substitute for the enacted statute, establish the contents of any future rule, or provide an official final agency disposition. Independent official rulemaking confirmation remains a reopening trigger.
Public Act 103-0769: Biometric Information Privacy Act amendment Illinois General Assembly · Primary Illinois enacted text
Does not eliminate BIPA duties, decide every pending case, or extend BIPA to every behavioral or mental-state inference.
Face Recognition Vendor Test Part 3: Demographic Effects (NISTIR 8280) National Institute of Standards and Technology · Primary NIST technical evaluation
Does not establish that every algorithm has identical error patterns, that identity matching reveals emotion or intent, or that laboratory results automatically predict every field deployment.
Accelerating AI Innovation Through Measurement Science National Institute of Standards and Technology · Primary NIST research and evaluation guidance
Does not validate any particular emotion, deception, loyalty, productivity, or vulnerability model and is not a certification of a deployed system.
ICO orders Serco Leisure to stop biometric employee-attendance monitoring UK Information Commissioner's Office · Primary data-protection enforcement record
Does not establish universal unlawfulness of all workplace biometrics, represent every worker's experience, or resolve law outside the United Kingdom.
Seldom Heard Voices: ethnic minority groups and gig economy workers' experiences UK Information Commissioner's Office / Thinks Insight & Strategy · Primary public research report with explicit qualitative sample limits
Does not provide a representative prevalence estimate for all ethnic-minority groups or gig workers, prove platform intent, or establish the outcome of a specific appeal.
iTutorGroup to pay $365,000 to settle EEOC discriminatory hiring suit U.S. Equal Employment Opportunity Commission · Primary agency litigation and settlement record
A settlement does not establish every alleged fact through trial, represent all automated hiring systems, or prove that every older applicant was affected in the same way.
Rite Aid facial-recognition case and modified order U.S. Federal Trade Commission · Primary agency enforcement record
Does not prove every allegation through a contested trial, establish the error rate of every face-recognition system, or extend the order beyond its parties and terms.
Examining the Examiners: Students' Privacy and Security Perceptions of Online Proctoring Services USENIX Symposium on Usable Privacy and Security · Primary research with disclosed sample and method
Does not represent all students, all disabilities, all proctoring products, or prove that every flagged event was erroneous or discriminatory.
Avoiding the Discriminatory Use of Artificial Intelligence U.S. Department of Education, Office for Civil Rights · Primary agency guidance; examples are illustrative rather than adjudicated findings
Does not adjudicate a specific school, replace statutory text, or prove that every example occurred in practice.
Reclaimed Term in Drag Performance Oversight Board · Independent platform oversight record; one illustrative appeal
Does not provide a platform-wide error rate, measure lost income, or establish that every reclaimed-term removal is wrongful.
Referring to Designated Dangerous Individuals as 'Shaheed' Oversight Board · Independent platform oversight analysis with stakeholder input
Does not bind all platforms, establish every removal's intent, or prove that every use of the term is benign.
Al-Shifa Hospital Oversight Board · Independent platform oversight record; one crisis-context case
Does not establish a universal platform pattern, determine every factual claim in the underlying conflict, or prove strategic effect from the removal.
User Experiences of Social Support From Companion Chatbots in Everyday Contexts Journal of Medical Internet Research / PubMed · Primary research using public reviews and a self-selected user sample
Does not establish clinical efficacy, long-term causal benefit, population prevalence, or safety for crisis or therapeutic use.
Overlapping Uses of AI Companions and AI Assistants ACM CHI Conference on Human Factors in Computing Systems · Primary HCI research with high-engagement, self-selected participants
Does not represent typical users, establish population prevalence, prove clinical benefit or harm, or resolve long-term dependency and disengagement outcomes.
Consumer Financial Protection Circular 2022-03: adverse action notification when creditors use complex algorithms Consumer Financial Protection Bureau · Primary agency interpretation of existing adverse-action duties
The circular was withdrawn on 2025-05-12, is not current CFPB guidance, does not govern every sector, and does not repeal or fully define the underlying statutory and regulatory duties.
Challenges to the Monitoring of Deployed AI Systems (NIST AI 800-4) National Institute of Standards and Technology · Primary NIST post-deployment evaluation guidance
Does not certify any particular system, define settled best practice for every sector, or prove that monitoring alone prevents harm.
Regulation (EU) 2026/1744 amending the AI Act and related regulations European Union / EUR-Lex · Primary enacted European Union legal text
Does not erase the AI Act, make all obligations immediately applicable, settle every exception, or supply legal advice for a particular system.
Code of Practice and Commission guidance on transparency of AI-generated content European Commission · Primary EU institutional implementation source
Does not prove compliance by any provider, make voluntary code participation universal, or establish the accuracy of a particular detection method.
Breast Cancer Awareness Content Oversight Board · Public independent oversight record with case-specific restoration outcome
Does not establish a platform-wide error rate, complete downstream reach repair, compensation, or long-term prevention of repeat errors.
Reporting on Somaliland Current Affairs Oversight Board · Public independent oversight record with page, post, and strike restoration
Does not establish complete repair of audience, income, reputation, or chilling effects, or a platform-wide prevalence rate for Somali-language enforcement error.
Comment on Kenyan Politics Using a Designated Slur Oversight Board · Public independent oversight record with policy-list currentness finding
Does not establish that every use of the term is harmless, that every language list is inaccurate, or that restoration repaired all prior visibility and participation effects.
Decision and Order in the Matter of IntelliVision Technologies Corp. U.S. Federal Trade Commission · Primary agency enforcement order
Does not independently validate the product, establish every alleged fact through contested trial, or convert testing documentation into field-validity certification.
Dispositivos neurotecnológicos: usos, regulación y antecedentes del caso Emotiv Insight Biblioteca del Congreso Nacional de Chile · Primary Chilean legislative research/currentness source
Does not establish comprehensive compliance, universal coverage of consumer neurotechnology, or that all ordered data deletion and downstream repair were independently verified.
Louis et al. v. SafeRent Solutions settlement administration record Court-authorized settlement administrator · Public settlement-administration record; settlement is not an admission of liability
Does not establish liability through trial, prove the validity or invalidity of every tenant-screening model, or show that all housing, credit, and downstream profile consequences were repaired.
Withdrawn Guidance — CFPB circulars and supervisory guidance Consumer Financial Protection Bureau · Primary agency source for withdrawal status
Does not repeal ECOA or Regulation B, decide the underlying statutory duties, or create a cross-sector explanation right.
Rite Aid: Modified Decision and Order U.S. Federal Trade Commission · Primary remedy and downstream-propagation source
Does not prove every third party completed deletion, every downstream copy was repaired, or every alleged harm was compensated.

Rights become real at the point of consequence and repair

How to read this research

Evidence before certainty

Inference before consequence

Remedy before legitimacy

Rights without impunity

Seven protections that work together

Freedom of thought

Mental privacy

Epistemic agency

Identity continuity

Right to know

Explanation and appeal

Freedom from engineered dependency

From the sovereign interior to public remedy

Foundations, Rights, and the Sovereign Interior

The Mind Is Not a Battlespace

Mental Privacy and AI Monitoring

The Invisible Editor and the Right to Know

Moderation, Suppression, and Appeal

Manifesto and Design Principles for a Free Mind

Civic Speech, Policy, and Local Action

Follow the chain from collection to remedy

Collection

Inference

Classification

Intervention

Consequence

Remedy

Different jurisdictions protect different parts of the problem

Experience is evidence—but an account is not automatically representative

A sensor reading is not a validated mental-state conclusion

Name the action before judging it

Removal

Access restriction

Age or region restriction

Search exclusion

Recommendation exclusion

Downranking or demotion

Reduced distribution

Demonetization or monetization change

Labeling or contextualization

Synthesized summaries and AI answers

Personalization

Account-level or content-level penalty

Automated refusal or response shaping

Saved-memory update, deletion, or profile change

A formal appeal is not meaningful unless it can repair the consequence

Timely, understandable notice

Access to relevant data and rule

Specific explanation

Correction of source and inferred data

Deletion and retention control

Authorized human review

Independent and discoverable appeal

Response time and interim protection

Restoration, compensation, and downstream repair

Audit logs and repeated-error prevention

Accessibility, language, and advocate support

Non-retaliation and safe correction

Aggregate public transparency

Withdrawal, exit, and relationship continuity

What changed after a challenge—and what remained unresolved

Enacted text, effective dates, guidance, inquiries, and rulings remain separate

Twelve declarations for a free mind

Seven exact owner reports and a bounded external registry

Exact owner-supplied reports

Reviewed primary and authoritative records