Descriptive mapping shows where recorded events occurred; it does not necessarily estimate the future.
PSYCHOLOGY / POLICING / GOVERNANCE
Psychological and “Pre-Crime” Initiatives
Behavioral threat assessment, predictive policing, watchlisting, traveler-risk analytics, behavioral science investigative units, and the governance of anticipatory state action.
This research atlas compares systems by what they actually do, whom or what they classify, what intervention follows, how long the designation persists, and whether correction or redress is possible. It does not treat every database, camera, map, behavioral team, or AI procurement as the same thing.
DIRECT ANSWER / EVIDENCE BOUNDARY
The collection in brief
- What is “pre-crime” here?
- A shorthand for anticipatory state decision architectures—not a claim that any system literally knows the future.
- Are threat assessment and predictive policing the same?
- No. Proper behavioral threat assessment is contextual, multidisciplinary, and management-oriented; predictive policing can forecast places or score people from historical data.
- What determines risk?
- The unit of analysis, evidence quality, intervention consequence, persistence, information sharing, feedback, and availability of correction or redress.
THE FULL SOCIOTECHNICAL CHAIN
The algorithm is only one link
A system becomes consequential through the interaction of data, identity, inference, human workflow, intervention policy, retention, and remedy. Select each step as a separate audit question.
-
01
Collection
Audit question: What data enters the system, under what authority, and with which omissions or distortions?
More data can widen coverage while also importing historic enforcement bias, administrative error, protected activity, and data collected for another purpose.
-
02
Identity resolution
Audit question: Which records are treated as belonging to the same person, place, device, network, or event?
A false match can aggregate unrelated facts into a persuasive dossier; a missed match can hide exculpatory or corrective information.
-
03
Inference or prioritization
Audit question: Is the system retrieving, classifying, forecasting, scoring, ranking, or merely displaying existing records?
Labels such as “decision support” can conceal materially predictive behavior; functional output and consequence matter more than product branding.
-
04
Human interpretation
Audit question: Does a person have time, evidence, training, authority, and a realistic ability to disagree?
A human review box does not cure opaque evidence, automation bias, queue pressure, or a decision space already narrowed by the system.
-
05
Intervention
Audit question: What changes because of the output: patrol allocation, support, surveillance, questioning, screening, detention, or denial?
The same score has different significance when it only informs aggregate planning versus when it creates persistent individual consequences.
-
06
Retention and sharing
Audit question: How long does the designation persist, who receives it, and can downstream users see its uncertainty and origin?
A temporary alert can become a durable identity attribute when copied across agencies, schools, employers, courts, or screening systems.
-
07
Feedback and model updating
Audit question: Does police attention generate the records later used to validate or retrain the system?
More patrol or scrutiny can produce more recorded incidents, contacts, and arrests, making deployment intensity look like predictive accuracy.
-
08
Review, correction, and redress
Audit question: Can an affected person discover, contest, correct, expire, or remove the designation and its downstream copies?
Without notice, audit logs, correction propagation, and independent review, errors can survive long after the original program changes name or ends.
FUNCTION BEFORE BRANDING
Distinctions that prevent category errors
The reports repeatedly warn that record retrieval, prediction, screening, assessment, surveillance, and coercive action are not interchangeable.
Place forecasting estimates where or when an event may occur; person-based scoring assigns consequences to named people and raises different due-process risks.
Identity resolution asks whether records refer to the same entity; it is not a prediction of future conduct.
A watchlist match is not proof of a crime, an arrest warrant, a probable-cause determination, or authority for every downstream action.
Behavioral threat assessment is a case-specific management process when properly practiced; it is not a published numerical probability that a person will attack.
Observable behavior, travel patterns, physiological arousal, and internal hostile intent are different constructs and cannot be treated as interchangeable.
A system can be technically sophisticated yet low-consequence, or computationally simple yet highly coercive. Consequence is not a proxy for model complexity.
Human review is meaningful only when reviewers have evidence, time, training, authority, and a functioning path to hold, reject, or correct the output.
Program discontinuation does not necessarily end the underlying function; watchlist matching, data fusion, or behavior observation may survive under a successor name.
A procurement notice, pilot, research paper, or vendor demonstration is not proof of a deployed operational system.
COMPARATIVE VIEW
Compare the unit and consequence before the model
| Dimension | Place forecasting | Person designation | Behavioral assessment | Screening / watchlisting |
|---|---|---|---|---|
| Unit of analysis | Area or time window | Named person or association | Referred case and contextual conduct | Identity, itinerary, or rule match |
| Typical output | Risk map or patrol area | Score, tier, list, or bulletin | Concern formulation and management plan | Match status or screening allocation |
| Primary evaluation question | Does it improve outcomes beyond transparent baselines and patrol dosage? | Does designation predict the stated outcome independently of increased scrutiny? | Does case management reduce risk without overlabeling protected or vulnerable people? | Are identity and rule matches accurate, proportionate, and correctable? |
| Characteristic rights risk | Unequal patrol burden and geographic feedback | Persistent individualized suspicion and coercive attention | Protected speech, disability, and weak-indicator overreach | Opaque travel, access, and identity consequences |
BEHAVIORAL SCIENCE / CASE MANAGEMENT / INVESTIGATIVE SUPPORT
Behavioral threat assessment and investigative units
The reports distinguish multidisciplinary threat assessment, behavioral investigative support, fusion-center reporting, and watchlist screening. A behavioral label does not establish a validated probability of violence.
Threat assessment and management
ActiveFBI Behavioral Threat Assessment Center (BTAC)
A national-level, multiagency, multidisciplinary center providing behavior-based operational support, research, training, consultation, and threat-management assistance.
Critical distinction: Public guidance describes structured professional judgment and case management, not an AI-generated probability that a person will attack.
Threat assessment and management
ActiveU.S. Secret Service National Threat Assessment Center (NTAC)
Conducts research, training, consultation, and operational guidance concerning targeted violence in protective, school, workplace, and community settings.
Critical distinction: Retrospective case patterns can guide inquiry but do not by themselves establish individual predictive validity.
Behavioral and investigative analysis
ActiveFBI Behavioral Analysis Units / National Center for the Analysis of Violent Crime
Specialized behavioral-science and criminal-investigative support functions that analyze cases, offender behavior, victimology, and investigative questions.
Critical distinction: Behavioral investigative analysis is not automatically a prospective risk score or threat-management decision.
Retained reports: [16]
Threat assessment and management
ActiveThreat Assessment and Threat Management (TATM) teams
Multidisciplinary local teams in schools, campuses, workplaces, health systems, communities, and law enforcement that receive and manage concrete concerns.
Critical distinction: Safeguards depend on referral thresholds, multidisciplinary evidence, services, retention, disability protections, and continuing reassessment.
Behavioral and investigative analysis
ActiveFusion centers and suspicious-activity reporting networks
Information-sharing and analytic centers that combine records, tips, intelligence, and partner reporting across jurisdictions.
Critical distinction: A broad collection threshold can turn protected or innocuous conduct into durable intelligence even without a predictive model.
Watchlisting and screening
ActiveFBI Threat Screening Center (TSC)
An interagency center responsible for threat-screening identity resolution, watchlist management, encounter coordination, and related information sharing.
Critical distinction: The center’s identity and screening functions should not be described as one self-executing predictive algorithm.
SYNTHETIC DECISION-ARCHITECTURE LAB
Decision Architecture and Feedback-Loop Lab
Select a synthetic scenario to see how data provenance, intervention intensity, feedback, evaluation, and redress change the meaning of a forecast.
Selected condition
Transparent hotspot baseline
A city uses recent reported incidents to identify small areas for a short, publicly defined patrol experiment and compares the result with ordinary hotspot mapping.
Data quality
Bounded
Target validity
Bounded
Intervention intensity
Medium
Feedback-loop risk
Medium
Independent evaluation
Strong
Notice and redress
Bounded
Analytical result
The forecast may be testable, but recorded incidents can still reflect patrol and reporting patterns. Evaluation must separate product value from the intervention.
Responsible next action
Pre-register outcomes, log patrol dosage, compare transparent baselines, and measure displacement and community burden.
COMPARATIVE PROGRAM REGISTER
Filter by function, unit, region, and status
The register includes operational programs, discontinued systems, pilots, enabling infrastructure, adjacent non-predictive systems, and conceptual proposals. Status and capability remain separate.
41 matching records
No program records match the current filters. Reset the filters or broaden the search.
Germany — Bavaria and Baden-Württemberg
PRECOBS
2013–present/locally reported
Near-repeat burglary forecasting from historical crime characteristics and trigger rules.
- Documented use
- Analyst-reviewed area alerts informed targeted patrol and prevention.
- Consequence
- Planning or resource allocation
- Evidence limit
- Independent evaluation found attribution of crime reduction unclear; public operational detail varies by locality.
- Rights-risk intensity
- Moderate
Retained reports: [6]
Germany — North Rhine-Westphalia
SKALA
2015–present/officially reported
Police-built place forecasting using crime records and contextual variables for burglary and vehicle crime.
- Documented use
- Risk maps informed shift and patrol deployment across district authorities.
- Consequence
- Planning or resource allocation
- Evidence limit
- Public evidence does not establish clear causal crime reduction or complete model details.
- Rights-risk intensity
- Moderate
Retained reports: [6]
Germany — Berlin
KrimPro
2016–2017
Pilot using burglary incidents and socioeconomic variables in 400-meter cells.
- Documented use
- Analyst-reviewed risk maps were tested for patrol use.
- Consequence
- Planning or resource allocation
- Evidence limit
- No public permanent continuation or independent impact evaluation was identified.
- Rights-risk intensity
- Moderate
Retained reports: [6]
Netherlands
Crime Anticipation System (CAS)
2015–2025
National multi-offense place forecasting system using police and contextual data.
- Documented use
- Generated geographic risk areas for operational planning.
- Consequence
- Planning or resource allocation
- Evidence limit
- The report says the system was phased out after moderate benefits; predictive validity and impact were not equivalent.
- Rights-risk intensity
- Moderate
Retained reports: [6]
Netherlands
System Risk Indication (SyRI)
2014–2020
Cross-government fraud-risk data integration often discussed as an adjacent anticipatory system.
- Documented use
- Generated risk indications for investigation rather than ordinary place patrol.
- Consequence
- Investigation or referral
- Evidence limit
- It was not a conventional police-crime forecast and was ended after legal challenge; included for governance comparison.
- Rights-risk intensity
- High
Retained reports: [6]
United Kingdom — Durham
Harm Assessment Risk Tool (HART)
2016–2021 publicly confirmed
Random-forest model classifying arrested people into two-year reoffending risk categories.
- Documented use
- Moderate-risk classification informed eligibility for the Checkpoint deferred-prosecution program, with officer discretion.
- Consequence
- Investigation or referral
- Evidence limit
- No sufficiently reliable public evidence in the report established continued operation after 2021; postcode variables and label design raised fairness questions.
- Rights-risk intensity
- High
Retained reports: [7]
United Kingdom
National Data Analytics Solution (NDAS)
2017–present/multi-use capability
Multi-use police analytics infrastructure with person-prediction research and investigative discovery workstreams.
- Documented use
- A serious-violence person model was not deployed after ethics review; modern-slavery work used NLP and network analysis.
- Consequence
- Investigation or referral
- Evidence limit
- NDAS is not one stable pre-crime algorithm; public status and use vary by workstream.
- Rights-risk intensity
- High
Retained reports: [7]
United Kingdom — London
Gangs Violence Matrix
2012–February 2024
Intelligence watchlist ranking named gang nominals using offenses, intelligence, victimization, and association.
- Documented use
- Inclusion affected targeting, stop and search, enforcement, safeguarding, and information sharing.
- Consequence
- Persistent police attention
- Evidence limit
- It was primarily a watchlist rather than machine-learning prediction; governance and disproportionality failures were documented.
- Rights-risk intensity
- Very high
United Kingdom — London
Violence Harm Assessment (VHA)
2024–present
Rules-based successor to the Gangs Violence Matrix with explicit thresholds and central governance.
- Documented use
- Used for quarterly harm assessment and operational prioritization under a published procedure.
- Consequence
- Persistent police attention
- Evidence limit
- The report notes continuing disproportionality and uncertainty about downstream effects despite governance changes.
- Rights-risk intensity
- Very high
Retained reports: [7]
United Kingdom — Kent
Kent PredPol deployment
2013–2018 reported
Operational test of proprietary short-term crime forecasting.
- Documented use
- Predicted place-time boxes informed patrol allocation.
- Consequence
- Planning or resource allocation
- Evidence limit
- Public evidence did not establish durable incremental benefit and the deployment ended.
- Rights-risk intensity
- Moderate
United Kingdom
National AI crime-mapping initiative
2025–2026 initiative
Emerging national effort to improve crime mapping and analytical support.
- Documented use
- Intended to support place analysis and allocation rather than a documented national person score.
- Consequence
- Planning or resource allocation
- Evidence limit
- Program design, validation, operational scope, and current deployment remain incomplete in public evidence.
- Rights-risk intensity
- Moderate
Retained reports: [7]
United States — Chicago
Strategic Subject List / CVRM
2012–2019
Models ranking people for future gun-violence involvement, combining victimization and offending risk.
- Documented use
- Selection produced custom notifications, police contact, and intervention priority.
- Consequence
- Persistent police attention
- Evidence limit
- Independent evaluation did not demonstrate violence reduction and documented increased police contacts.
- Rights-risk intensity
- Very high
United States — Los Angeles
LAPD LASER chronic-offender program
2011–2019
Manually administered points and chronic-offender bulletins supported by analytical platforms.
- Documented use
- Bulletins directed officers to locate, contact, investigate, and collect additional information.
- Consequence
- Persistent police attention
- Evidence limit
- Selection, removal, documentation, and causal outcome evidence were weak or inconsistent.
- Rights-risk intensity
- Very high
United States — Pasco County, Florida
Intelligence-Led Policing program
2011–successor practices
Person and youth lists presented as identifying prolific or at-risk people.
- Documented use
- Selected people and families received repeated visits, monitoring, warrant checks, code enforcement, and heightened accountability.
- Consequence
- Persistent police attention
- Evidence limit
- No public evaluation established predictive validity of the original lists; operational attention itself generated confirming records.
- Rights-risk intensity
- Very high
United States and other deployments
PredPol / Geolitica
2011–2024 company closure
Commercial self-exciting place-time forecasting product deployed in multiple cities.
- Documented use
- Generated short-term patrol boxes for selected offenses.
- Consequence
- Planning or resource allocation
- Evidence limit
- Vendor claims and favorable experiments do not establish consistent incremental crime reduction; company operations ended.
- Rights-risk intensity
- Moderate
United States — multiple cities
HunchLab / Missions
2010s–successor products
Commercial place forecasting and patrol-mission allocation combining crime and contextual data.
- Documented use
- Produced risk surfaces and mission recommendations for patrol.
- Consequence
- Planning or resource allocation
- Evidence limit
- Public code, independent field evaluation, and current product lineage vary; forecasting and optimization must be assessed separately.
- Rights-risk intensity
- Moderate
Retained reports: [4]
Multiple jurisdictions
Risk Terrain Modeling deployments
various
Environmental analysis relating land uses and facilities to crime risk.
- Documented use
- Used diagnostically or prospectively to focus place-based prevention and patrol.
- Consequence
- Planning or resource allocation
- Evidence limit
- A static diagnostic map is not necessarily a future forecast; implementations and evidence vary.
- Rights-risk intensity
- Moderate
Retained reports: [4]
United States
FBI BTAC / TATM Initiative
2010–present publicly described
National multidisciplinary behavioral threat assessment, consultation, training, and team-building capability.
- Documented use
- Supports case inquiry, investigative strategy, management planning, and reassessment.
- Consequence
- Support, management, or safeguarding
- Evidence limit
- No standard national algorithmic attack-probability formula is publicly documented; effectiveness evidence is constrained by rare events and referred-case selection.
- Rights-risk intensity
- High
United States
Secret Service NTAC
1998–present
Research, guidance, training, and consultation for targeted-violence prevention.
- Documented use
- Promotes multidisciplinary reporting, information gathering, individualized management, and follow-up.
- Consequence
- Support, management, or safeguarding
- Evidence limit
- NTAC rejects a demographic attacker profile; public guidance does not provide a national numerical probability instrument.
- Rights-risk intensity
- High
United States — trained law-enforcement users
NCBIO-25 structured aid
current use reported
Human-administered structured-professional-judgment aid for threat-assessment personnel.
- Documented use
- Organizes investigative inquiry and concern formulation.
- Consequence
- Support, management, or safeguarding
- Evidence limit
- It is not AI, a population-screening instrument, or a validated probability calculator.
- Rights-risk intensity
- High
United States
Threat Screening Center / TSDS
continuing
Federated watchlisting enterprise for terrorism identities within a broader threat-screening environment.
- Documented use
- Performs nomination review, identity resolution, designation, encounter coordination, modification, removal, and redress support.
- Consequence
- Travel or access restriction
- Evidence limit
- Public records omit sensitive thresholds and complete performance metrics; the watchlist itself does not supply every downstream legal authority.
- Rights-risk intensity
- Very high
United States
TSC AI Enhancement Initiative (FBI-TSC-AIE)
2026 sources-sought notice
Market research for federated discovery, identifier correlation, source-linked synthesis, and visualization.
- Documented use
- Documented prediction language concerned where additional relevant information may be found across systems.
- Consequence
- Research or conceptual output
- Evidence limit
- No public award, person-risk score, automated watchlisting decision, or operational deployment was established by the report cutoff.
- Rights-risk intensity
- High
Retained reports: [14]
United States — TSA
Secure Flight
2009–present
Identity resolution and allocation of boarding or checkpoint screening status from passenger data and lists.
- Documented use
- Matches traveler identity and communicates screening outcomes.
- Consequence
- Enhanced screening or observation
- Evidence limit
- Public precision, false-match, and subset details are limited by security restrictions.
- Rights-risk intensity
- Very high
United States — CBP
ATS / Unified Passenger
continuing
Border decision-support environment combining manifests, reservations, prior encounters, watchlists, intelligence, and rules.
- Documented use
- Prioritizes travelers for inspection or review.
- Consequence
- Enhanced screening or observation
- Evidence limit
- Rules, weighting, error rates, and complete downstream actions are not publicly measurable.
- Rights-risk intensity
- Very high
United States — TSA
SPOT / integrated behavior detection
2003–successor practice
Operational behavior-observation program using trained indicators and referrals.
- Documented use
- Officers referred travelers for additional screening or law-enforcement checks.
- Consequence
- Enhanced screening or observation
- Evidence limit
- Scientific evidence did not establish reliable aviation-threat detection; behavior-detection functions survived dedicated workforce changes.
- Rights-risk intensity
- Very high
Retained reports: [12]
United States — DHS research
FAST / Passive Methods research
2007–public lineage unclear
Research attempting to infer mal-intent from remotely sensed physiological and behavioral signals.
- Documented use
- Laboratory and planned field-assessment work, not documented routine passenger decisions.
- Consequence
- Research or conceptual output
- Evidence limit
- Balanced mock-event accuracy and prototypes do not establish real-world base rates, construct validity, or operational deployment.
- Rights-risk intensity
- Very high
Retained reports: [12]
United States — TSA/FAMS
Quiet Skies / Silent Partner
2010–June 2025
Rules-based selection of some non-watchlisted travelers for enhanced screening and, in some cases, air-marshal observation.
- Documented use
- Matched travel patterns and rules to allocate screening and surveillance resources.
- Consequence
- Enhanced screening or observation
- Evidence limit
- TSA did not establish comprehensive effectiveness measures; termination did not end all underlying vetting functions.
- Rights-risk intensity
- Very high
Retained reports: [12]
China — Xinjiang
Integrated Joint Operations Platform (IJOP)
documented 2010s; current configuration unclear
Region-specific platform aggregating records, checkpoints, sensors, household information, and rule-based warnings.
- Documented use
- Generated investigation tasks and flags concerning named people, relationships, places, and behavior.
- Consequence
- Detention or coercive intervention
- Evidence limit
- Public evidence is granular about an associated mobile application but not complete server logic, current configuration, thresholds, or all downstream decisions.
- Rights-risk intensity
- Extreme
China — provincial and municipal
Police Cloud / public-security big-data platforms
2010s–present procurement and publicity
Provincial or municipal infrastructures consolidating police and administrative records for search, relationship reconstruction, warnings, and analysis.
- Documented use
- Support identity, network, location, anomaly, and investigative workflows.
- Consequence
- Investigation or referral
- Evidence limit
- Procurement describes intended capability, not uniform national delivery, validation, or use.
- Rights-risk intensity
- Very high
China — urban/local systems
Skynet video networks
various
Label generally used for urban public-security video networks and associated recognition or command functions.
- Documented use
- Supports observation, identification, vehicle/person matching, and response.
- Consequence
- Investigation or referral
- Evidence limit
- The label varies across place and time and should not be treated as one predictive national platform.
- Rights-risk intensity
- Very high
China — county, township, village networks
Sharp Eyes
2010s–present/local variation
Video networking and community-governance initiative extending coverage beyond major urban systems.
- Documented use
- Connects community and non-police cameras with public-security and social-governance workflows.
- Consequence
- Investigation or referral
- Evidence limit
- Interoperability does not prove one database, model, or command system; local functions vary.
- Rights-risk intensity
- Very high
India
CCTNS / ICJS
national rollout continuing
National criminal-record and criminal-justice interoperability infrastructure linking police and justice-sector records.
- Documented use
- Supports search, sharing, investigation, and potential analytics across institutions.
- Consequence
- Investigation or referral
- Evidence limit
- Coverage counts do not establish data quality, lawful reuse, predictive function, or every connected application’s behavior.
- Rights-risk intensity
- High
Retained reports: [9]
India — Delhi
Crime Mapping, Analytics and Predictive System (CMAPS)
2014–implementation gap documented through 2019
Planned crime mapping, threat rating, network analysis, profiling, and open-source modules.
- Documented use
- Verified operational core was closer to mapping and descriptive decision support than a mature validated predictor.
- Consequence
- Planning or resource allocation
- Evidence limit
- Auditor findings documented incomplete integration and unimplemented advanced modules.
- Rights-risk intensity
- High
Retained reports: [9]
India — Telangana
TSCOP
documented existence; details limited
Police mobile and analytical platform connecting records, field checks, and identification workflows.
- Documented use
- Supports search, verification, and operational police information access.
- Consequence
- Investigation or referral
- Evidence limit
- Public documentation does not establish a validated person-level future-offending model or complete feature set.
- Rights-risk intensity
- High
Retained reports: [9]
India — Uttar Pradesh
Trinetra / YAKSH
documented deployment/publicity
Facial-recognition and record-linking tools used to identify or investigate people.
- Documented use
- Supports biometric matching and analytical access to police records.
- Consequence
- Investigation or referral
- Evidence limit
- Identification is not future-conduct prediction; public accuracy, false-match, and governance data are limited.
- Rights-risk intensity
- Very high
Retained reports: [9]
India — Maharashtra
MARVEL / MahaCrimeOS products
2020s initiatives
State-linked analytics and AI product initiatives for police records, investigation, and decision support.
- Documented use
- Public descriptions include search, analysis, linking, and workflow support.
- Consequence
- Investigation or referral
- Evidence limit
- Existence is documented, but complete production features, prediction targets, validation, and independent audit are not public.
- Rights-risk intensity
- High
Retained reports: [9]
Australia — New South Wales
Suspect Targeting Management Plan (STMP)
about 2000–December 2023
Long-running preventive-policing program selecting people believed likely to reoffend for individualized management.
- Documented use
- Plans could direct stops, searches, visits, bail checks, monitoring, warnings, arrest, referrals, and coordination.
- Consequence
- Persistent police attention
- Evidence limit
- Oversight found systemic documentation, youth, legality, and disproportionality problems; model and discretion changed over time.
- Rights-risk intensity
- Very high
Australia — New South Wales
Prevention, Disruption, Response (PDR)
2024–present successor
Successor issue-management procedures after STMP discontinuation.
- Documented use
- Police describe data-informed local tasking and management.
- Consequence
- Persistent police attention
- Evidence limit
- Exact predictive capability, internal compliance review, selection, and effects remain incompletely public.
- Rights-risk intensity
- Very high
Retained reports: [8]
Australia — Queensland
Queensland place-forecasting research
research studies
Research comparing short-term forecasts for burglary and vehicle-related theft across local areas.
- Documented use
- Evaluated forecast performance, not the causal effect of operational patrol deployment.
- Consequence
- Research or conceptual output
- Evidence limit
- Optimal parameters varied by place and offense; research does not establish statewide operational use.
- Rights-risk intensity
- Moderate
Retained reports: [8]
United States and other jurisdictions
Acoustic gunshot detection
various
Sensors classify and locate suspected gunfire after a sound event.
- Documented use
- Generates dispatch alerts and incident records.
- Consequence
- Investigation or referral
- Evidence limit
- It is event detection, not future prediction, though it can enter feedback loops or be bundled with predictive products.
- Rights-risk intensity
- High
Retained reports: [4]
Multiple jurisdictions
Real-Time Crime Centers
various
Centers integrating dispatch, cameras, license-plate readers, records, alerts, and analyst communications.
- Documented use
- Supports real-time response and investigation.
- Consequence
- Investigation or referral
- Evidence limit
- Not inherently predictive; classification depends on whether the center generates or operationalizes future-risk estimates.
- Rights-risk intensity
- High
TEN RESEARCH ROUTES
Follow the system from definition to remedy
Each route preserves source qualifications, distinguishes operational status from possibility, and names the questions the public record does not settle.
Decision Architecture and Definitions
Why “pre-crime” is better understood as a chain of collection, inference, intervention, retention, and redress than as one algorithm.
Open research routeBehavioral Threat Assessment and Management
BTAC, NTAC, multidisciplinary teams, structured professional judgment, supportive intervention, and the line where prevention becomes coercive labeling.
Open research routePlace-Based Forecasting and Patrol
Hot spots, near repeats, risk terrain, patrol optimization, vendor claims, dosage, displacement, and feedback.
Open research routePerson-Based Risk Lists and Targeting
Chicago SSL, LAPD LASER, Pasco County, HART, London’s former Matrix, Australia’s STMP, and the consequences of a durable designation.
Open research routeWatchlisting, Traveler Risk, and Mal-Intent
Nomination, identity resolution, screening allocation, travel-pattern rules, behavior observation, and research attempting to infer hostile intent.
Open research routeGlobal Comparative Systems
Europe, the United Kingdom, United States, China, India, and Australia—compared by function, unit, status, and consequence rather than by the label “AI police.”
Open research routeTechnology Stack, Feedback, and Data
The full lifecycle from sensors and records through identity resolution, features, models, dashboards, action, feedback, correction, deletion, and retirement.
Open research routeLaw, Governance, Remedy, and Audit
AI Act boundaries, data protection, due process, equality, notice, procurement transparency, model evaluation, appeals, correction, and retirement.
Open research routeHuman–Machine Judgment, Bias, and Provenance
Automation bias, triage traps, explanation laundering, hidden alternatives, source lineage, and deterministic policy control.
Open research routeAnticipatory Intelligence and Warfare Boundaries
Conceptual GAITE research, physical-change forecasting, federated analysis, legal review, and the distinction between population preemption and theater-level situational forecasting.
Open research routeGOVERNANCE / RIGHTS / ASSURANCE
Ten controls for a consequential anticipatory system
Purpose and legal authority
State the precise purpose, statutory authority, prohibited uses, decision owner, and downstream actions before collecting or scoring data.
Data provenance and necessity
Use only lawful, necessary, relevant data; document origin, collection context, quality, known bias, corrections, and restrictions on reuse.
Identity and association safeguards
Separate confirmed identity, alias, association, co-location, victimization, witness status, and allegation; do not let weak linkage become a stable person attribute.
Independent validation
Publish target definitions, denominators, error measures, calibration, subgroup performance, baselines, field-test design, and causal outcome limits.
Meaningful human judgment
Reviewers need evidence, uncertainty, contrary information, time, training, authority, and a real ability to hold, reject, or correct an output.
Proportional intervention
Match intervention intensity to verified facts and lawful authority; a probabilistic flag alone should not authorize coercive individual action.
Feedback-loop controls
Track whether deployment creates the contacts, arrests, reports, or detections later used as evidence of accuracy.
Notice, reasons, and challenge
Provide timely notice and understandable reasons where lawful; create accessible review, correction, appeal, and representation paths.
Retention, sharing, and correction propagation
Set expiration and deletion rules, record every disclosure, and propagate corrections or removals to all recipients and derived products.
Retirement and successor audit
Ending a product name is not enough. Verify shutdown, archive lawful evidence, delete unlawful or obsolete data, and audit successor workflows for inherited practices.
RETAINED RESEARCH LIBRARY
Twenty reports preserved as source inputs
These cards identify the report basis for the public synthesis. They do not expose repository paths and do not convert every embedded statement into verified site fact.
Submitted Comparative Report
The Real-World “Pre-Crime” Landscape: A Comparative Inventory and Governance Framework for Predictive Law Enforcement
Comparative inventory covering 36 programs in 15 countries. It supplies the collection-wide decision-architecture model, intensity distinctions, and governance framework.
Submitted Research Plan
The Global Evolution of Predictive and Preemptive Law Enforcement
Broad global research plan and taxonomy. Its stronger claims remain research leads unless corroborated elsewhere in the retained library.
Submitted Technical Report
The Global Technology Stack Behind Predictive Law Enforcement
Lifecycle and technology-stack synthesis spanning collection, matching, feature construction, scoring, intervention, feedback, correction, deletion, and retirement.
Submitted Technical Audit
Hot Spots, Feedback Loops, and Vendor Claims: A Technical Audit of Place-Based Predictive Policing
Technical audit distinguishing descriptive mapping, short-term forecasting, patrol optimization, risk-terrain modeling, event detection, and person-based prediction.
Submitted Case Study
Hot People: Chicago’s Strategic Subject List, LAPD LASER, and Pasco County’s Intelligence-Led Policing Program
Comparative audit of three U.S. person-based systems and the operational consequences of designation, contact, and feedback loops.
Submitted Legal Analysis
Predictive Policing after the EU AI Act: PRECOBS in Germany, CAS in the Netherlands, and the Boundary Between Prohibited and High-Risk AI
European systems analysis and an AI Act decision tree separating place forecasting from prohibited or high-risk person assessment.
Submitted Country Report
Predictive Policing in the United Kingdom: NDAS, HART, the Gangs Violence Matrix, and AI Crime Mapping
United Kingdom program history covering person prediction, analytical infrastructure, intelligence matrices, successor governance, and place mapping.
Submitted Country Report
Policing Future Risk in Australia: The NSW Suspect Targeting Management Plan, Operation Tepito, and Successor Practices
Australia case study centered on STMP, youth and Indigenous impact, police powers, oversight, discontinuation, and successor practices.
Submitted Country Report
Predictive Policing in India: From CMAPS and CCTNS to State-Level AI Risk and Surveillance Systems
India systems inventory distinguishing national record infrastructure, mapping, biometric identification, state analytics, and unverified predictive claims.
Submitted Country Report
Predictive Public Security in China: IJOP, Police Cloud, Skynet, Sharp Eyes, and the Expansion of Proactive Policing
Country report on layered public-security infrastructure, IJOP, Police Cloud, video networks, proactive warnings, and consequences.
Submitted Country Report
Predictive Public Security in China: IJOP, Police Cloud, Skynet, Sharp Eyes, and the Expansion of Proactive Policing — Part 2
Second independent treatment of the Chinese ecosystem, emphasizing functional distinctions, coercive intervention, and evidence limits.
Submitted Program Audit
Detecting Mal-Intent Before an Attack: FAST, SPOT, Quiet Skies, Secure Flight, and U.S. Traveler-Risk Analytics
Aviation and border analysis separating identity matching, travel-pattern rules, observable behavior, and attempted inference of internal intent.
Submitted Process Analysis
From Intelligence Nomination to Government Action: How the U.S. Terrorism-Watchlisting System Works and Where Predictive AI Could Enter It
Process map of nomination, identity review, screening distribution, agency action, encounter feedback, correction, and redress.
Submitted Procurement Analysis
The FBI Threat Screening Center AI Enhancement Initiative: Is It a Predictive “Pre-Crime” System?
Procurement-status analysis concluding that the documented initiative is market research for federated discovery and analyst support, not a deployed person-risk model.
Submitted Behavioral Report
Behavioral Threat Assessment or Pre-Crime? The FBI BTAC, Secret Service NTAC, and the Prevention of Targeted Violence
Evidence-qualified analysis of structured professional judgment, multidisciplinary management, base rates, civil rights, and the boundary between prevention and pre-crime.
Submitted Behavioral Report
Federal Threat Assessment Programs Evaluation
A second report on BTAC, NTAC, fusion centers, schools, structured instruments, supportive and punitive interventions, and safeguards.
Submitted Architecture Report
AI Bias Mitigation in Intelligence
Conceptual work on automation bias, command compression, the triage trap, provenance, competing hypotheses, and deterministic control layers.
Submitted Concept Report
Secure Federated Intelligence Architecture
Conceptual decentralized intelligence architecture. It is retained as a proposal and threat-modeling source, not evidence of a deployed GAITE system.
Submitted Feasibility Framework
Anticipatory Intelligence Engine Research
Feasibility and evaluation framework for a conceptual anticipatory intelligence engine, including forecasting, provenance, privacy, adversarial testing, and legal review.
Submitted Ethical Analysis
Anticipatory Warfare Ethical Analysis
Conceptual ethical comparison between individual behavioral prediction and theater-level analysis of physical changes. It does not establish legality or deployment of a GAITE system.
PREDICTION INSIDE GOVERNMENT
Decision support becomes governance when it changes access, priority, inspection, or intervention.
Municipal and national proxy governance can operationalize predictive systems; the authority and redress chain matters as much as the model.
Trace proxy responsibility, review, and remedy Follow material change, correction, and retirementSHARED TEST
A human review label is not enough.
Both collections test evidence access, decision authority, review time, intervention capacity, appeal, correction, and whether a system quietly pre-structures the available options.
FROM PRE-CRIME TO ANTICIPATORY WARFARE
The same word “prediction” can conceal radically different units and consequences.
Compare individualized pre-crime with theater-level warning and military target-decision systems.
The connected atlas compares individualized rare-event prediction with theater-level warning, predictive target nomination, autonomous kill chains, and kill webs while preserving the legal and evidentiary boundary between them.