Machine-speed decisions need machine-held evidence. Bounded education only: no real targets, operational control, executable payloads, or live-system actions. A human click is not a liability transfer. For production evidence and accountability software, visit Evulgare.com ↗.
Curated public record
Source library
The library favors official policy, official program documentation, technical standards, primary framework publications, and organization-controlled product material. Manufacturer claims are labeled and not treated as independent performance verification.
129 source recordsReview date: 2026-08-07External links open separately
The site prefers official law, policy, doctrine, standards, procurement and program records, then clearly attributed manufacturer material, independent research, and high-quality reporting.
It supports the bounded statement described in the source record—not every claim on the destination page or every configuration and later version of a system.
The Source Freshness and Correction Center shows the current editorial review state, scheduled review date, typed public dependencies, supersession, and correction pathway for every source ID.
Bounded support: Company-described loitering-system sensing, mission control, tracking, and post-launch functions; not a universal statement of engagement authority.
Bounded support: Company-described endpoint, identity, cloud, threat-intelligence, investigation, and response capabilities used in security operations.
Bounded support: Company-described radar, electro-optical, and sensor technologies for surveillance, tracking, fire control, and situational awareness.
Bounded support: Company-reported delivery of uncrewed ground vehicles for operational support roles; not independent verification of every configuration.
Bounded support: Company announcement of Army prototype selection for AI/ML-enabled deep sensing and decision support; developmental status remains material.
Bounded support: Public description of the interactive kill-web explorer, observed release characteristics, direct answers, feature routes, synthetic boundary, and sister-site relationship.
Bounded support: KillWebs.com working definition of a governed option space, selective reachability, later binding, and the relation between a web and a selected mission chain.
Bounded support: Structural comparison of selected chains and wider kill-web option spaces, including topology, authority, data, testing, failure response, and the roles of both sister sites.
Bounded support: Public description of a synthetic, non-operational capability-marketplace exercise that separates technical feasibility, constraints, comparison, authority, and execution.
Bounded support: Public-safe organization of kill-web concepts, programs, architecture layers, governance questions, evidence roles, research IDs, and public unknowns without tactical locations.
Bounded support: Public description of a synthetic lab examining evidence, time, authority, intervention, communications, lifecycle decisions, and reconstructable accountability.
Bounded support: Public description of sensing, correlation, option composition, policy and trust checks, authorization, assessment, and controlled recomposition.
Bounded support: Public explanation of capability semantics, data fabrics, transport, cloud-edge placement, identity, trust, orchestration, interfaces, and graceful degradation.
Bounded support: Public treatment of integrity, timing, identity, data poisoning, supply-chain, electromagnetic, and common-mode risks without operational attack procedures.
Bounded support: Historical description of PRECOBS as a near-repeat, place-based burglary forecasting approach and its analyst-mediated operational workflow. It does not establish current deployment or independent crime-reduction effects.
Bounded support: Historical official evidence that PRECOBS remained in use in the Nuremberg policing area in 2019. It does not establish the 2026 deployment footprint.
Bounded support: CAS purpose, national grid-cell design, twelve-week data window, human use of outputs, operational dates, and out-of-use status from December 2025. It does not establish a causal crime-reduction effect.
Bounded support: Public-record handling and documented governance questions concerning CAS data, evaluation, procedures, retention, and impact assessments. It does not by itself prove which requested records existed or how every local team used CAS.
Bounded support: SSL/CVRM purpose, operational and governance weaknesses, evaluation limitations, and CPD decommissioning of the Party to Violence models on November 1, 2019.
Bounded support: Development of HART with Durham Constabulary, its training data, random-forest classification, risk categories, and advisory use in custody and Checkpoint eligibility. It does not establish current use in 2026.
Bounded support: Public-sector description of HART as a supervised machine-learning tool used by Durham Police and discussion of bias, governance, and human decision-making. It does not establish current status or complete subgroup performance.
Bounded support: Retrospective public account that HART was used from 2016 through 2021. It does not establish whether HART remained operational after that period.
Bounded support: Official Met account of the former Gangs Violence Matrix and its decommissioning. It does not independently validate the Matrix’s selection, effects, or fairness.
Bounded support: Regulatory findings concerning the former Matrix’s governance, accuracy, retention, differentiation, and information sharing. It does not characterize every individual entry or later VHA practice.
Bounded support: VHA purpose, inclusion and scoring approach, published DPIA/EIA/SOP, annual MOPAC review, access/removal information, and current operational status. It does not establish causal effectiveness.
Bounded support: Current VHA population and demographic figures as of July 3, 2026, including 1,632 people meeting criteria. It does not establish that inclusion is accurate or effective for every person.
Bounded support: Secure Flight’s identity data, prescreening workflow, system purpose, records, sharing, and privacy controls. It does not disclose all matching rules or aggregate error rates.
Bounded support: Current TSA passenger-screening process and public travel guidance. It does not independently validate Secure Flight matching accuracy or disclose controlled criteria.
Bounded support: Public watchlisting lifecycle, nomination and identity-resolution roles, screening uses, and the distinction between a watchlist record and an agency’s downstream legal action.
Bounded support: The March 2025 renaming of the Terrorist Screening Center to the Threat Screening Center and the publicly described expansion of its mission. It does not disclose all dataset criteria or operating rules.
Bounded support: Watchlist scale, review and redress processes, accuracy and re-review concerns, and recommendations. Aggregate counts do not constitute a false-positive rate.
Bounded support: Current FBI BTAC and TATM operational support, multidisciplinary staffing, training, research, and behavior-based threat-assessment role. It does not establish a national AI future-danger score.
Bounded support: Multidisciplinary unit design, case-specific behavioral threat assessment, management, and continuing prevention practice. It does not establish that one indicator or profile predicts attack.
Bounded support: Evulgare identity, live domain, accountability principles, evidence-layer framing, and its public description of the KillChains–KillWebs–Evulgare ecosystem. It does not independently verify a customer deployment, certification, legal conclusion, or operational performance.
Bounded support: Evulgare’s public platform model connecting evidence, observation, transformation, model, uncertainty, interpretation, policy, authority, interface, human judgment, decision, outcome, provenance, and assurance. It does not independently establish production integration or performance.
Bounded support: The public product-area boundary for technical and causal attribution across evidence, model, software, policy, authority, interface, human review, and organizational controls. It explicitly does not claim to decide legal liability.
Bounded support: The public provenance model for append-oriented histories connecting entities, activities, agents, transformations, versions, authority, interface state, and later invalidations. It does not independently establish a deployed integration.
Bounded support: The public assurance model connecting claims, evidence, assumptions, counterclaims, defeaters, reviewers, configurations, and invalidating changes. It does not certify any external system or claim.
Bounded support: The public distinction among technical capability, trusted data, delegated authority, permitted purpose, and contextual permission. It does not establish the actual authority state of a third-party system.
Bounded support: The public human-judgment model preserving raw evidence, contrary indicators, alternatives, review time, abstention, and independent hypothesis formation. It does not prove that any particular interface provides meaningful control.
Bounded support: The public uncertainty model separating aleatoric and epistemic uncertainty, distribution status, evidence completeness, source dependence, and model disagreement. It does not validate any external model’s uncertainty estimates.
Bounded support: The public trust model treating identity, data integrity, semantic compatibility, disclosure, permitted use, and authority as separate dimensions. It does not prove that a connected source is trustworthy or authorized.
Bounded support: The public resilience model for partitions, shared dependencies, stale evidence, integrity failures, safe holds, quarantine, and reconciliation. It does not establish the resilience of a deployed third-party system.
Bounded support: The public change-impact model for tracing which claims, evidence, assumptions, interfaces, and verification activities are affected by software, model, dataset, policy, compiler, or environment changes. It does not automatically invalidate or revalidate a real system.
A source link supports the bounded statement described in its “supports” field. It does not automatically support every claim on the destination page, every configuration of a product, or every later version of a program. Product and program pages can change after publication; dates and archived release artifacts matter.
The source library is a curated evidence set, not a claim that all relevant public material has been discovered. Classified records, proprietary implementation details, operator doctrine, test evidence, and incident data are frequently unavailable.
The public source-index.json file mirrors this registry for discovery clients. It is derived metadata: this human-readable page and the linked original sources remain the evidence path.