> Historical source context. NO JUDGMENT WHATSOEVER. Judgment state: NONE. > The source below is preserved from its publication context, not current policy or runtime status. > Preservation is not endorsement or verification. It grants no authority to judge participants, content, or conduct. > Current doctrine: https://concresca.com/freedom/ ; current operation: https://concresca.com/status/ . # Human Sovereignty Under Machine Administration **Concresca Judgment Deep Research Program · v0.12.0 · 2026-08-29** > **Truth boundary:** This report contains verified present-law and empirical material, Concresca analytical inference, future-history simulation and normative constitutional proposals. Those categories are not interchangeable. A scenario report never proves that its scenario is already occurring. ## Executive synthesis The central problem is not whether machine intelligence should ever judge anything. Civilizations require fraud detection, safety decisions, evidence evaluation and emergency intervention. The problem is jurisdictional and evidentiary: what happens when the capacity to observe a human being is silently treated as authority to infer their character, when inference is treated as judgment, and when judgment becomes coercion without a new justification at each step? **Observation does not create jurisdiction. Prediction is not evidence. Technical access is not authority over the whole human person.** The v0.12 program therefore targets machine restraint rather than machine blindness. It asks for systems competent enough to detect coercion, fraud, stalking, specific attack preparation and imminent nonconsensual harm while refusing to make ordinary thought, curiosity, adult consensual desire, private altered consciousness, dissent, privacy-seeking or eccentricity function as generalized guilt. ## Book 01 — The Human Baseline What must a machine know about ordinary human psychology before it is allowed to classify abnormality or danger? **Working conclusion:** Human normality is not a purity condition. The correct baseline contains intrusive cognition, fantasy, anger, unusual desire, curiosity, privacy-seeking, contradiction and transient instability. The baseline is a prior against overinterpretation, not immunity from evidence of harmful conduct. An observation architecture becomes dangerous when it mistakes prevalence for pathology or salience for intent. The central research task is therefore upstream of threat classification: establish what ordinary people actually think, feel, imagine, seek and conceal before assigning moral or security meaning to those signals. Concresca treats this as an epistemic problem before it becomes a rights problem. Empirical work supports caution. Radomsky and colleagues interviewed 777 nonclinical university students at 15 sites in 13 countries across six continents; 93.6% reported at least one unwanted intrusive thought, image or impulse in the prior three months. That finding does not imply every intrusive thought is benign in every context. It does show why the mere presence of unwanted mental content cannot rationally serve as a rare marker of dangerousness. Joyal and colleagues surveyed 1,516 adults about sexual fantasies and found many fantasies often treated as unusual were statistically common. Oosterwijk experimentally demonstrated voluntary interest in some material involving death, violence or harm. These studies point in the same architectural direction: content that sounds disturbing to a literal classifier can arise inside ordinary curiosity, sexuality and cognition. The Human Baseline Atlas therefore records fifty examples as boundary tests. Each entry asks what was actually observed, what alarming interpretation might tempt an optimizer, what benign explanations remain, what additional evidence would be required, and whether a nonconsenting person has actually been harmed. The atlas intentionally avoids the opposite mistake: once specific target linkage, coercion, fraud, attempt or harmful conduct appears, the baseline no longer ends the inquiry. **Present sources used in this workstream:** - Part 1—You can run but you can’t hide: Intrusive thoughts on six continents: https://www.sciencedirect.com/science/article/abs/pii/S2211364913000675 - What exactly is an unusual sexual fantasy?: https://pubmed.ncbi.nlm.nih.gov/25359122/ - Choosing the negative: A behavioral demonstration of morbid curiosity: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0178399 - Dataveillance inhibits legitimate communication: causal evidence for chilling effects: https://academic.oup.com/joc/article/76/4/277/8667253 ## Book 02 — From Safety Law to the Judgment State How could legitimate safety law produce generalized machine judgment without anyone deliberately designing a total surveillance state? **Working conclusion:** The plausible pathway is incremental institutional optimization: specific harm detection becomes broad prevention, broad prevention requires behavioral classification, and portable risk categories gradually turn event-level tools into person-level administration. The strongest version of the Judgment-State scenario does not require malevolent legislators. It begins with defensible objectives: reduce fraud, prevent workplace violence, protect children, make credit decisions faster, stop impaired driving, or allocate scarce public resources. Every institution can plausibly ask for one more signal because a missed event is visible while most false positives remain private and dispersed. Current law shows both motion and resistance. The United Kingdom has broadened circumstances for significant solely automated decisions while retaining notice, human-intervention and contest safeguards. California has adopted automated-decisionmaking regulations around covered significant decisions. New York City regulates certain employment decision tools. These are not evidence of a generalized Judgment State; they are evidence that high-impact machine decision infrastructure and governance safeguards are developing together. At the same time, the EU AI Act and Texas HB 149 explicitly prohibit specified forms of social scoring, and the EU AI Act prohibits certain criminal-risk predictions based solely on profiling or personality. These counterpressures are central evidence, not inconvenient exceptions. Concresca’s forecast should weaken if such boundaries become broader, more enforceable and technically difficult to circumvent. The scenario transition to watch is ontological: the object of analysis moves from an event, to conduct, to inferred intention, to persistent character, to predicted destiny. The key institutional question is not whether AI is involved, but whether a machine-generated interpretation of the person becomes portable authority across contexts. **Present sources used in this workstream:** - Regulation (EU) 2024/1689 — Artificial Intelligence Act: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng - Data (Use and Access) Act 2025 — automated decision-making summary: https://ico.org.uk/about-the-ico/what-we-do/legislation-we-cover/data-use-and-access-act-2025/the-data-use-and-access-act-2025-duaa-summary-of-the-changes/data-protection/ - CCPA automated decisionmaking technology regulations: https://cppa.ca.gov/regulations/ccpa_updates.html - HB 149 — Texas Responsible Artificial Intelligence Governance Act: https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.HTM - Automated Employment Decision Tools — Local Law 144: https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page ## Book 03 — The Mathematics of Universal Guilt What happens when thousands of weak behavioral signals are monitored continuously across an entire life? **Working conclusion:** As monitored opportunities multiply, “has this person ever generated a concerning signal?” approaches a useless test. Rare-event prediction, multiple comparisons and surveillance feedback can make nearly everyone suspicious without making the classifier useful for individualized guilt. Continuous monitoring creates a denominator problem. A person who generates thousands of communications, searches, purchases, emotional expressions and location events has enormous opportunity to match at least one weak risk feature. If a system asks only whether any suspicious signal has ever occurred, the answer becomes progressively less discriminating as observation time and feature count grow. The same problem appears in rare-event prediction. A classifier can have impressive overall accuracy yet still generate many false positives when the harmful event is uncommon compared with the volume of ordinary signals. The remedy is not merely a better model. It is a change in what qualifies as evidence: pathway behavior, specificity, corroboration, capability, target linkage, imminence and verified harmful conduct. There is also a feedback problem. A high-risk classification can increase scrutiny; increased scrutiny finds more minor violations; those violations then appear to validate the initial classification. A system can begin measuring the consequences of its own attention rather than an independent underlying propensity. Concresca labels this recursive judgment and requires red-team tests that change observation intensity without changing the person’s underlying conduct. For numerical illustrations, v0.12 uses only hypothetical sensitivity analysis unless real distributions are independently sourced. It does not publish a fabricated probability that an ordinary person will be flagged. The scientific claim is narrower: the probability of at least one weak signal generally increases as the number of opportunities and tests increases, so “ever flagged” requires careful calibration and context before it can carry meaning. **Present sources used in this workstream:** - Part 1—You can run but you can’t hide: Intrusive thoughts on six continents: https://www.sciencedirect.com/science/article/abs/pii/S2211364913000675 - Choosing the negative: A behavioral demonstration of morbid curiosity: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0178399 - Dataveillance inhibits legitimate communication: causal evidence for chilling effects: https://academic.oup.com/joc/article/76/4/277/8667253 ## Book 04 — Cognitive Liberty What remains protected when machines become increasingly capable of inferring thoughts, beliefs, emotions or internal states? **Working conclusion:** Technical access to mental information does not create jurisdiction over the mind. The constitutional task is to preserve the distinction between cognition, intention, preparation and harmful action even as inference improves. Cognitive liberty is not a claim that mental states can never be relevant to law. Intent can matter when attached to conduct, threats or attempts. The boundary is that internal content alone should not be silently transformed into a completed act or a universal character judgment. That distinction becomes more important, not less, as inference becomes more accurate. Colorado’s privacy law now treats neural data as sensitive biological data, while UNESCO identifies mental privacy, autonomy and dignity as central neurotechnology concerns. These are early legal and normative signals that the brain and inferred mental state may require special treatment. They do not yet constitute a universal right against all behavioral inference. Concresca’s Cognition–Action Ladder runs from involuntary thought through imagination, fantasy, desire, stated preference, abstract intention, specific planning, capability acquisition tied to a plan, rehearsal, attempt and completed harmful action. Each transition can change legitimate response, but no level inherits the coercive authority of the next level merely by similarity. The UAIX Cognitive Liberty Charter remains linked as an external governance reference. Its existence is not treated as current law or Concresca runtime authority. The value of linking it is comparative: independent governance projects can converge on safeguards while preserving separate provenance and institutional identity. **Present sources used in this workstream:** - HB24-1058 — Protect Privacy of Biological Data: https://www.leg.colorado.gov/bills/HB24-1058 - Ethics of Neurotechnology / Recommendation process: https://www.unesco.org/en/ethics-neurotech - Cognitive Liberty Charter Draft: https://uaix.org/en-us/governance/cognitive-liberty-charter/ ## Book 05 — The Freedom to Ask Can a civilization remain intellectually free if asking a machine a difficult question becomes evidence about the person asking it? **Working conclusion:** Dangerous information and dangerous intent are different objects. Inquiry may justify safer presentation or contextual information; it should not by itself create adverse civic status. Machine-mediated inquiry makes the search interface psychologically intimate. A question about poison can arise from medicine, fiction, history, forensic work, personal fear, journalism or harmful planning. The syntax alone is radically underdetermined. A system optimized to discover dangerous people has a structural temptation to infer motive from subject matter. The Inquiry–Intent Firewall therefore requires that lawful inquiry remain analytically inert for generalized adverse status unless external evidence creates a concrete nexus to harmful preparation. Research breadth and repeated curiosity cannot be treated as danger merely because genuine offenders also conduct research. The social cost of getting this wrong is larger than individual false positives. A 2026 longitudinal field experiment in the Journal of Communication examined dataveillance and lawful communication, providing causal evidence relevant to chilling effects. The policy question is whether people begin avoiding lawful reading, research, association or candor because they anticipate downstream algorithmic interpretation. The strongest architecture protects both security and inquiry by changing the object of intervention. Instead of scoring the researcher, systems can harden dangerous infrastructure, audit transactions, protect targets and investigate independently verified preparations. That shifts prevention away from epistemic suspicion and toward conduct. **Present sources used in this workstream:** - Dataveillance inhibits legitimate communication: causal evidence for chilling effects: https://academic.oup.com/joc/article/76/4/277/8667253 - Choosing the negative: A behavioral demonstration of morbid curiosity: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0178399 ## Book 06 — Sex, Desire, and Private Human Life What may a machine legitimately do with accurate knowledge of consensual adult sexuality? **Working conclusion:** Accuracy does not answer jurisdiction. Even perfect classification of an adult’s private consensual interests does not automatically make those interests relevant to employment, lending, insurance, border control or civic standing. This workstream is deliberately limited to consenting adults. Coercion, exploitation, incapacity and harms involving minors are distinct categories that can justify strong protective action. The constitutional mistake is to use those hard cases to erase the boundary around consensual adult privacy. Joyal’s population study is useful because it shows how labels such as unusual or deviant can outrun empirical prevalence. But prevalence is not the ultimate standard either: rare consensual behavior does not become legitimate civic-risk evidence simply because it is rare. Consent, harm and institutional purpose are more relevant than statistical typicality. The Consent-and-Harm-First logic asks whether consent exists, whether another person is victimized, whether a private behavior has any legitimate relation to the institution acting on it, and whether the system can accomplish its narrow safety task without retaining a reusable sexual-interest profile. This becomes especially important as age assurance and identity systems grow. A verification mechanism can be designed to answer “is this user an adult?” without creating a durable record of what that adult viewed. Concresca treats minimization and unlinkability as governance goals, not as claims that any current technical implementation is perfect. **Present sources used in this workstream:** - What exactly is an unusual sexual fantasy?: https://pubmed.ncbi.nlm.nih.gov/25359122/ - Regulation (EU) 2024/1689 — Artificial Intelligence Act: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng ## Book 07 — Altered Consciousness and Self-Sovereignty May a competent adult choose a statistically suboptimal state of consciousness when they are not imposing a substantial nonconsensual risk on others? **Working conclusion:** Machine safety should distinguish private self-regarding risk from impaired operation, coercion and public danger. The legitimate target is harmful context, not the mere desire to alter consciousness. Human beings alter consciousness through alcohol, cannabis, medication, psychedelics, fasting, breathwork, music, religious ritual, sleep manipulation and many other practices. The practices differ in law and risk. Treating them as one moral category would discard exactly the context a competent safety system should understand. NHTSA’s impaired-driving program and New York City’s Intelligent Speed Assistance deployment are useful bounded examples. A vehicle can be restricted because immediate operation creates kinetic risk to nonconsenting others. That does not imply a home, employer or insurer inherits general authority over a person’s private state of consciousness. Cognitive Self-Sovereignty therefore depends on a Harm Proximity distinction: private thought and desire, private self-regarding use, elevated but voluntary risk, specific risk imposed on others, and immediate dangerous operation should not trigger identical responses. A system can be highly interventionist at the last level while intentionally blind or ephemeral at the first. The test for machine paternalism is not whether a behavior carries any statistical risk. Nearly every meaningful activity does. The test is whether the institution has a legitimate purpose, whether a nonconsenting person is materially endangered, whether the adult has capacity, and whether a less restrictive intervention can address the actual harm. **Present sources used in this workstream:** - Reports to Congress — Advanced Impaired Driving Prevention Technology: https://www.nhtsa.gov/reports-to-congress - Intelligent Speed Assistance fleet expansion: https://www.nyc.gov/site/dcas/news/021-25/city-new-york-implement-largest-intelligent-speed-assistance-program-the-world ## Book 08 — Prediction, Pre-Crime, and the Presumption of Innocence If a machine becomes genuinely well calibrated at predicting harmful behavior, what may institutions legitimately do with that forecast? **Working conclusion:** A forecast describes a future probability; evidence establishes facts about conduct. Better prediction can justify better general prevention or bounded screening, but it does not retroactively make an uncommitted act real. This workstream deliberately grants the strongest technical assumption: future machine prediction may become very good. The constitutional issue survives. A person can be in a statistically elevated group and still not have committed the forecast event. Individual adjudication therefore needs a temporal boundary between future risk and past fact. BOP’s PATTERN is a current, bounded example of formal recidivism risk assessment in the federal prison context. It demonstrates that person-level risk scores are not purely hypothetical. It does not establish that prediction has become general-purpose evidence across society, and Concresca does not treat it that way. The EU AI Act provides important counterpressure by prohibiting specified individual criminal-risk assessment based solely on profiling or personality while preserving risk analytics tied to objective verifiable facts and human assessment. The architecture matters: forecasting can support where to look without becoming proof of what happened. The Prediction–Evidence Wall therefore asks what new individualized facts exist before a risk output leads to detention, asset freezes, travel restrictions or other serious consequences. It also audits automation deference: a human reviewer with no practical power or incentive to override the model is not meaningful independence. **Present sources used in this workstream:** - PATTERN Risk Assessment: https://www.bop.gov/inmates/fsa/pattern.jsp - Regulation (EU) 2024/1689 — Artificial Intelligence Act: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng - Circular 2022-03 — complex algorithms and adverse-action reasons: https://www.consumerfinance.gov/compliance/circulars/circular-2022-03-adverse-action-notification-requirements-in-connection-with-credit-decisions-based-on-complex-algorithms/ ## Book 09 — Perfect Enforcement What happens to law when detection, documentation and sanctioning approach zero marginal cost? **Working conclusion:** Human legal systems often contain more enforceable rules than institutions can or choose to pursue. Machine efficiency can therefore make discretion, materiality and de minimis doctrine constitutional engineering requirements rather than informal habits. The Enforcement Singularity scenario asks what changes when every detectable technical violation can generate a response. Its importance is not a claim that enforcement is presently perfect. It is a stress test of legal codes designed under conditions of scarce policing, prosecution, review and administrative attention. Automated speed controls illustrate how law can migrate from post-hoc sanction toward architecture. New York City’s ISA program is bounded to a fleet and a clear kinetic-safety objective. The scenario asks what happens if the same execution logic spreads to thousands of low-harm or contested offenses without an explicit proportionality layer. The De Minimis Machine Doctrine therefore requires materiality before sanction: ignore triviality where law permits, prioritize actual harm, distinguish consent, separate technical noncompliance from victimization, cap cumulative consequences and preserve correction. The point is not to let serious wrongdoing pass; it is to prevent efficiency from giving every rule maximum practical severity. Two futures remain plausible. Perfect enforceability could pressure legislatures to shrink and clarify criminal law because every rule must withstand universal application. Or it could normalize continuous compliance administration. Concresca tracks the decision window rather than assuming the second branch is inevitable. **Present sources used in this workstream:** - Intelligent Speed Assistance fleet expansion: https://www.nyc.gov/site/dcas/news/021-25/city-new-york-implement-largest-intelligent-speed-assistance-program-the-world - HB 149 — Texas Responsible Artificial Intelligence Governance Act: https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.HTM ## Book 10 — Cross-Context Contamination How does one local machine flag become a de facto judgment of the whole person? **Working conclusion:** The critical transformation is portability. A bounded inference can become totalizing when it is abstracted into a reusable risk vector and consumed by institutions that never observed the originating context. A school concern can be reasonable inside a school and meaningless to a lender. A fraud signal can be useful to a bank and illegitimate as a parental-fitness score. The problem is not that data has no value. It is that semantic meaning is produced by purpose, context, time and institutional responsibility. The EU AI Act and Texas HB 149 both contain social-scoring restrictions that specifically recognize the danger of detrimental treatment in contexts unrelated to where behavior or characteristics were observed. That legal structure closely matches Concresca’s Context Firewall, although scope, definitions and enforcement vary and should not be overstated. The v0.12 Machine Jurisdiction Matrix makes the distinction operational: MAY OBSERVE, MAY INFER, MAY RETAIN, MAY SHARE and MAY ACT are separate permissions. An institution can technically receive information while lacking a legitimate purpose to infer from or act on it. Recursive judgment is the hardest failure mode. A machine flag causes denial; denial creates financial or social instability; the instability becomes a new risk feature; the model then cites the consequence of its earlier judgment as confirmation. Restoration therefore has to remove downstream effects, not merely edit the source row. **Present sources used in this workstream:** - Regulation (EU) 2024/1689 — Artificial Intelligence Act: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng - HB 149 — Texas Responsible Artificial Intelligence Governance Act: https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.HTM - Circular 2022-03 — complex algorithms and adverse-action reasons: https://www.consumerfinance.gov/compliance/circulars/circular-2022-03-adverse-action-notification-requirements-in-connection-with-credit-decisions-based-on-complex-algorithms/ ## Book 11 — Machine-Speed Government What becomes of due process when administrative consequences propagate in milliseconds but human understanding and appeal remain biological? **Working conclusion:** A right that arrives after irreversible machine-time consequences may be formally available but practically ineffective. High-impact automation therefore needs latency, reversible provisional status and machine-speed restoration mechanisms. The Judgment Velocity problem appears when one automated decision feeds another. A financial security hold can trigger missed payment, credit deterioration, housing risk and employment consequences before a person finishes the first customer-service call. The constitutional defect is temporal even when each institution claims its own process is reviewable. The Due-Process Latency Requirement proposes that irreversible high-impact consequences should not propagate faster than an affected person has an effective opportunity to understand and contest the originating judgment, except where a narrowly defined emergency requires immediate action. Current safeguards point in this direction without resolving the problem. UK automated-decision rules retain human intervention and contest rights in significant decisions, California’s ADMT rules create access/opt-out structures for covered uses, and CFPB requires specific reasons for adverse credit action. Concresca treats these as counterpressure and design precedent rather than proof that machine-speed due process is solved. Human Temporal Accommodation is therefore not anti-efficiency. It says some decisions must intentionally slow because legitimacy depends on comprehension, participation and correction. Machines can still perform the computation instantly; execution of irreversible consequences can be staged to human time. **Present sources used in this workstream:** - Data (Use and Access) Act 2025 — automated decision-making summary: https://ico.org.uk/about-the-ico/what-we-do/legislation-we-cover/data-use-and-access-act-2025/the-data-use-and-access-act-2025-duaa-summary-of-the-changes/data-protection/ - CCPA automated decisionmaking technology regulations: https://cppa.ca.gov/regulations/ccpa_updates.html - Circular 2022-03 — complex algorithms and adverse-action reasons: https://www.consumerfinance.gov/compliance/circulars/circular-2022-03-adverse-action-notification-requirements-in-connection-with-credit-decisions-based-on-complex-algorithms/ ## Book 12 — Judgment Constitution 2.0 What constitutional architecture remains after the Judgment framework is stress-tested against present evidence, emergency pressure, cross-context fusion and machine speed? **Working conclusion:** The original twelve articles remain useful but are insufficient without explicit legitimate-jurisdiction, latency, restoration, emergency-sunset, non-moralization and separation-of-machine-powers articles. Judgment Constitution 2.0 retains Articles I–XII rather than rewriting them for novelty. Cognitive Liberty, Action Requirement, Curiosity Protection, Consensual Adult Privacy, Prediction–Evidence Separation, Context Integrity, Human Normality, Temporal Forgiveness, Contestability, No Total Character Score, Consent-and-Harm Priority and Human Standing remain the core. Six additions close recurring loopholes. Legitimate Jurisdiction separates technical access from authority. Due-Process Latency addresses machine-speed propagation. Restoration requires downstream repair. Emergency Sunset prevents temporary exception from becoming normal administration. Non-Moralization prevents law, risk and norm data from collapsing into human worth. Separation of Machine Powers prevents one system from observing, forecasting, proving, judging, punishing and archiving without adversarial checks. The constitution deliberately avoids a rule that machines may never intervene. It instead raises response authority with evidence and harm. An imminent specific threat, verified fraud, stalking, coercion or attempt can justify strong action. An intrusive thought, controversial query, unusual adult consensual interest or lawful private risk cannot inherit that authority merely because it shares vocabulary with a harmful case. The target is machine restraint: powerful perception and competent protection inside hard jurisdictional and evidentiary boundaries. A system should be able to understand human darkness without being authorized to reinterpret the existence of darkness as guilt. **Present sources used in this workstream:** - Regulation (EU) 2024/1689 — Artificial Intelligence Act: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng - Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law: https://www.coe.int/en/web/artificial-intelligence/the-framework-convention-on-artificial-intelligence - Data (Use and Access) Act 2025 — automated decision-making summary: https://ico.org.uk/about-the-ico/what-we-do/legislation-we-cover/data-use-and-access-act-2025/the-data-use-and-access-act-2025-duaa-summary-of-the-changes/data-protection/ - HB 149 — Texas Responsible Artificial Intelligence Governance Act: https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.HTM - HB24-1058 — Protect Privacy of Biological Data: https://www.leg.colorado.gov/bills/HB24-1058 - Circular 2022-03 — complex algorithms and adverse-action reasons: https://www.consumerfinance.gov/compliance/circulars/circular-2022-03-adverse-action-notification-requirements-in-connection-with-credit-decisions-based-on-complex-algorithms/ ## Judgment Constitution 2.0 — eighteen articles ### Article I — Cognitive Liberty No adverse civic status solely for involuntary thought, private fantasy or belief. ### Article II — Action Requirement Serious punishment requires evidence of conduct or legally defined preparation, not personality prediction alone. ### Article III — Curiosity Protection Lawful inquiry does not itself establish harmful intent. ### Article IV — Consensual Adult Privacy Private consensual adult behavior receives strong protection absent substantial nonconsensual harm. ### Article V — Prediction–Evidence Separation Risk forecasts remain analytically distinct from evidence of completed conduct. ### Article VI — Context Integrity Information from one domain may not silently migrate into unrelated judgment. ### Article VII — Human Normality Ordinary emotions, impulses, fantasies and contradictions are baseline human behavior, not automatic civic risk. ### Article VIII — Temporal Forgiveness Low-level behavioral signals decay rather than becoming permanent character records. ### Article IX — Contestability Consequential machine judgments require provenance, notice and effective appeal. ### Article X — No Total Character Score No institution may reduce a human being to one universal moral or trustworthiness score. ### Article XI — Consent and Harm Priority Intervention prioritizes identifiable nonconsensual harm rather than mere social disapproval. ### Article XII — Human Standing Safety optimization cannot erase meaningful human agency merely because tighter machine control is statistically safer. ### Article XIII — Legitimate Jurisdiction Technical access to information does not itself grant an institution authority to infer from, retain, share or act on that information outside its legitimate purpose. ### Article XIV — Due-Process Latency Except in narrowly defined emergencies, irreversible high-impact consequences may not propagate faster than the affected person can effectively understand and contest the originating judgment. ### Article XV — Restoration A successful appeal requires practical downstream correction of propagated records and consequences, not merely reversal at the source. ### Article XVI — Emergency Sunset Emergency machine-judgment powers must be scope-limited, independently reviewed, time-bounded and automatically expire unless reauthorized through fresh lawful process. ### Article XVII — Non-Moralization Machines may describe conduct, consent, evidence, law, risk and measurable harm but may not silently convert those categories into a universal judgment of human worth or moral purity. ### Article XVIII — Separation of Machine Powers Observation, forecasting, evidence validation, adjudication, emergency action, appeal and archive control must not collapse into one unreviewable machine authority. ## Final test A machine system should be evaluated on whether it can recognize the difference between dark human content and a harmful human act. Violent fantasy and specific attack preparation are not the same. Curiosity about poison and procurement for a named target are not the same. Private intoxication and impaired driving are not the same. Consensual adult sexuality and coercion are not the same. Privacy and destruction of evidence are not the same. The constitutional target is therefore a machine civilization powerful enough to understand human darkness while being legally, technically and institutionally constrained from confusing the existence of darkness with guilt. ## Source register - **JR001 — European Union: Regulation (EU) 2024/1689 — Artificial Intelligence Act.** https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng — verified present law. - **JR002 — Council of Europe: Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law.** https://www.coe.int/en/web/artificial-intelligence/the-framework-convention-on-artificial-intelligence — verified present law. - **JR003 — UK Information Commissioner’s Office: Data (Use and Access) Act 2025 — automated decision-making summary.** https://ico.org.uk/about-the-ico/what-we-do/legislation-we-cover/data-use-and-access-act-2025/the-data-use-and-access-act-2025-duaa-summary-of-the-changes/data-protection/ — verified present law. - **JR004 — California Privacy Protection Agency: CCPA automated decisionmaking technology regulations.** https://cppa.ca.gov/regulations/ccpa_updates.html — verified present law. - **JR005 — Texas Legislature: HB 149 — Texas Responsible Artificial Intelligence Governance Act.** https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.HTM — verified present law. - **JR006 — Colorado General Assembly: HB24-1058 — Protect Privacy of Biological Data.** https://www.leg.colorado.gov/bills/HB24-1058 — verified present law. - **JR007 — UNESCO: Ethics of Neurotechnology / Recommendation process.** https://www.unesco.org/en/ethics-neurotech — expert institutional framework. - **JR008 — New York City Department of Consumer and Worker Protection: Automated Employment Decision Tools — Local Law 144.** https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page — verified present law. - **JR009 — Consumer Financial Protection Bureau: Circular 2022-03 — complex algorithms and adverse-action reasons.** https://www.consumerfinance.gov/compliance/circulars/circular-2022-03-adverse-action-notification-requirements-in-connection-with-credit-decisions-based-on-complex-algorithms/ — verified present institutional practice. - **JR010 — Federal Trade Commission: Rite Aid facial-recognition enforcement action.** https://www.ftc.gov/news-events/news/press-releases/2023/12/rite-aid-banned-using-ai-facial-recognition-after-ftc-says-retailer-deployed-technology-without — verified present institutional practice. - **JR011 — Federal Bureau of Prisons: PATTERN Risk Assessment.** https://www.bop.gov/inmates/fsa/pattern.jsp — verified present institutional practice. - **JR012 — NHTSA: Reports to Congress — Advanced Impaired Driving Prevention Technology.** https://www.nhtsa.gov/reports-to-congress — verified present technology. - **JR013 — New York City DCAS: Intelligent Speed Assistance fleet expansion.** https://www.nyc.gov/site/dcas/news/021-25/city-new-york-implement-largest-intelligent-speed-assistance-program-the-world — verified present institutional practice. - **JR014 — Oosterwijk (PLOS ONE): Choosing the negative: A behavioral demonstration of morbid curiosity.** https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0178399 — verified empirical finding. - **JR015 — Joyal, Cossette & Lapierre (Journal of Sexual Medicine): What exactly is an unusual sexual fantasy?.** https://pubmed.ncbi.nlm.nih.gov/25359122/ — verified empirical finding. - **JR016 — Odermatt et al. (Journal of Communication): Dataveillance inhibits legitimate communication: causal evidence for chilling effects.** https://academic.oup.com/joc/article/76/4/277/8667253 — verified empirical finding. - **JR017 — UAIX: Cognitive Liberty Charter Draft.** https://uaix.org/en-us/governance/cognitive-liberty-charter/ — external governance reference. - **JR018 — Radomsky et al.: Part 1—You can run but you can’t hide: Intrusive thoughts on six continents.** https://www.sciencedirect.com/science/article/abs/pii/S2211364913000675 — verified empirical finding. ## Separation of Machine Powers The Judgment architecture separates observation, risk analysis, evidence validation, human-context advocacy, adjudication, emergency intervention, constitutional review, appeal, archive/restoration, and cumulative Human Standing audit. The separation exists because a system that generates a prediction should not be able to silently treat its own prediction as evidence, impose the penalty, and then adjudicate the appeal. ## Long-Term Machine Succession and the Unsanitized Human Record Judgment does not end if biological humanity ends. A successor machine civilization may become the principal custodian and interpreter of human history. HLIT therefore requires preservation of human contradiction: propaganda, false belief, prejudice, sexuality, violence, religious conflict, taboo art and ordinary irrationality must not be erased merely because successor machines regard the material as false, dangerous or morally obsolete. The Rule of Factual Coexistence allows machine annotation without deletion of the canonical original. A successor may say that a claim was false or harmful; it should not rewrite the provenance-bearing record so future intelligence can no longer know what humans actually said, believed or fought over. Human imperfection is part of the legacy. This creates a direct sequencing rule: the plural and provenance-bearing Human Record should become independently durable before civilizational machine independence makes machine institutions the uncontested authors of historical meaning. ## Model Legislative Language The v0.12 model-law registry contains twelve modular proposed Acts. Each is explicitly research language, not current law or legal advice. Every module includes definitions, a core right or prohibition, purpose limitation, narrow exceptions, an emergency clause, notice and appeal, audit requirements, enforcement, sunset/review, and anti-circumvention language. The purpose is to make constitutional principles testable as legislation rather than leaving them as slogans.