Evidence Atlas · Release 1.2.0

Explore the public record without flattening uncertainty.

Name: Global Kill Chain Evidence Atlas
Creator: KillChains.com

Move from a country-level overview through doctrine, named systems, kill-chain functions, human control, deployment status, evidence, timeline, governance, and source records. The map is deliberately abstract: it contains no tactical deployment points, active-unit locations, or targeting coordinates.

9 country / alliance records12 system / program records13 visible evidence statesCountry-level abstraction only

10interconnected research layers

9country, partnership, or alliance entries

12named system, program, or policy records

11dated publication and framework events

Progressive disclosure

Ten linked layers preserve context from map to source.

The Atlas is an evidentiary graph rather than a capability leaderboard. Each layer answers a different question and makes it harder for a visually polished claim to appear more certain than its source permits.

01 Country State or alliance-level entry point connecting public posture, programs, systems, and sources.
02 Doctrine & policy Published strategy, policy, public limits, and declared operating principles.
03 System Platform, software, sensor, command architecture, program, or other named capability.
04 Kill-chain function The exact Find, Fix, Track, Target, Engage, or Assess function supported or automated.
05 Human control The publicly described distribution of authority, supervision, veto, abort, and safe-state behavior.
06 Deployment status Research, demonstrated, developmental, fielded, mixed, or publicly unspecified status.
07 Evidence & confidence Visible source class, claim state, confidence, disputes, and unresolved unknowns.
08 Timeline Exact or bounded dates for policy, announcement, demonstration, fielding, reporting, correction, or review.
09 Legal & governance Public review rules, international processes, declared principles, and limitations.
10 Source The traceable public record supporting the claim and its evidence classification.

World overview

Filter the evidence before interpreting the map.

Search by country, doctrine, system, function, human-control arrangement, evidence state, or lifecycle status. “Verified only” retains official, court/treaty, and independently corroborated primary states.

Search the Atlas

Region

Kill-chain function

Human control

Primary evidence

Lifecycle status

Verified primary states only

9 visible records

Country-level abstraction

Public evidence map

9 / 9

Abstract position only · no operational coordinates

Verified primary state Qualified claim Disputed or unresolved

Country record · North America Reviewed 2026-07-31

United States

Officially documented High for named programs; many operational details remain classified

Public U.S. evidence is strongest for AI-assisted intelligence, sensor fusion, target cueing, command-and-control, autonomy research, and bounded autonomous functions—not a blanket transfer of lethal authority to AI.

Public doctrine / posture: Public policy requires appropriate levels of human judgment, realistic testing, safety, cybersecurity, understandable interfaces, and legal compliance. Named programs span decision support, autonomy research, and bounded system functions.
Kill-chain functions: Find · Fix · Track · Target · Engage · Assess
Human control: Mixed or system-specific
DoD Directive 3000.09 requires appropriate levels of human judgment, realistic testing against adaptive adversaries, safety and cybersecurity, understandable interfaces, and legal compliance.
Lifecycle status: Mixed lifecycle status across programs or configurations
Governance context: DoD Directive 3000.09 is the clearest public control record in this dataset. It does not make every U.S. program or configuration equivalent.
Unknowns and limits: Classified rules, exact model architectures, operational thresholds, and weapon-release modes for many systems are not public.

Open table record

SourcesU.S. Department of Defense 2023-01-25 U.S. Chief Digital and Artificial Intelligence Office 2026-07-23 Defense Innovation Unit 2024

Country record · Middle East Reviewed 2026-07-31

Israel

Manufacturer-described High for advertised functions; independent operational verification varies

Israeli manufacturers publicly describe systems spanning automatic target recognition, human-in-the-loop precision weapons, and bounded autonomous anti-radiation loitering munitions.

Public doctrine / posture: Publicly available material in this release is system- and manufacturer-centered rather than a complete national doctrine record. Control arrangements must be assessed by function, configuration, and mode.
Kill-chain functions: Find · Fix · Track · Target · Engage · Assess
Human control: Mixed or system-specific
Control arrangements differ by system and mode; “autonomous” marketing language must be parsed at the function–context level.
Lifecycle status: Mixed lifecycle status across programs or configurations
Governance context: The Atlas does not infer a single national autonomy policy from product pages or conflict reporting. It preserves disputed and unknown fields rather than filling them.
Unknowns and limits: Operational software, thresholds, deployment rules, review procedures, and many conflict-specific claims are not fully public or independently verifiable.

Open table record

SourcesIsrael Aerospace Industries Current product page Israel Aerospace Industries Current product page Rafael Advanced Defense Systems Current product page

Country record · Europe Reviewed 2026-07-31

Norway

Manufacturer-described High for fielding and advertised terminal-recognition functions

Norway’s clearest public example is the Naval Strike Missile family, where autonomous target recognition supports terminal discrimination after a human-defined mission.

Public doctrine / posture: The clearest public record in this release concerns bounded post-launch sensing, navigation, and terminal recognition within a human-defined mission.
Kill-chain functions: Find · Fix · Track · Engage
Human control: Bounded post-activation autonomy
Public material establishes mission planning and terminal recognition but does not fully disclose intervention and release logic for every operator and configuration.
Lifecycle status: Fielded or operationally introduced
Governance context: Public product material does not disclose every operator-specific release, intervention, software, classifier, or abort rule.
Unknowns and limits: Exact software, classifier performance, abort behavior, and country-specific rules of engagement are not public.

Open table record

SourcesKongsberg Defence & Aerospace Current product page

Country record · East Asia Reviewed 2026-07-31

People’s Republic of China

Independently corroborated Moderate for strategic direction; lower for system-by-system deployment claims

PLA writings and external research emphasize “intelligentization,” human–machine integration, distributed sensing, decision advantage, swarms, and kill-web concepts. Public operational details are limited.

Public doctrine / posture: External research describes PLA interest in intelligentization, human–machine integration, multi-domain sensing, decision advantage, swarms, and kill-web concepts.
Kill-chain functions: Find · Fix · Track · Target · Engage · Assess
Human control: Publicly unspecified
Publicly accessible governance and weapon-release rules are less transparent than those published by the United States or NATO.
Lifecycle status: Developmental or research program
Governance context: The public record in this release supports strategic direction more strongly than system-by-system fielding, control mode, or operational validation.
Unknowns and limits: The maturity, combat validation, control modes, and fielded scale of particular AI-enabled kill-chain functions are often unknown.

Open table record

SourcesRAND Corporation 2025 National Defense University Press 2025 CNA 2026-07

Country record · Europe / Eurasia Reviewed 2026-07-31

Russia

Disputed Moderate; conflict claims require caution

Russia publicly promotes AI-assisted drone recognition, terminal autonomy, electronic-warfare resilience, and reconnaissance–strike integration. The exact degree of autonomous target selection is frequently unclear.

Public doctrine / posture: Public claims and independent analysis describe AI-assisted recognition, electronic-warfare resilience, and reconnaissance–strike integration, while precise autonomy and control arrangements remain contested or undisclosed.
Kill-chain functions: Find · Fix · Track · Engage
Human control: Publicly unspecified
Public manufacturer language does not reliably establish who authorizes each engagement, how abort works, or how models behave outside advertised conditions.
Lifecycle status: Mixed lifecycle status across programs or configurations
Governance context: Manufacturer and belligerent claims are visually demoted and kept separate from independent analysis. The Atlas does not convert a product claim into verified combat performance.
Unknowns and limits: Combat performance, error rates, software versions, operator oversight, and autonomy modes are contested or undisclosed.

Open table record

SourcesZALA Aero 2026-07-23 Center for Strategic and International Studies 2026

Country record · Europe Reviewed 2026-07-31

Ukraine

Credibly reported Moderate; fast iteration creates version and terminology uncertainty

Ukraine has rapidly adopted AI-assisted detection, navigation, targeting support, and terminal guidance to cope with electronic warfare and the scale of drone operations.

Public doctrine / posture: Public reporting describes rapid adoption of AI-assisted detection, navigation, targeting support, terminal guidance, imagery triage, and defensive uses under intense electronic-warfare pressure.
Kill-chain functions: Find · Fix · Track · Engage · Assess
Human control: Publicly unspecified
Different systems may be remotely piloted, supervised, or terminally autonomous; “AI-enabled” should not be read as proof of fully autonomous weapon release.
Lifecycle status: Mixed lifecycle status across programs or configurations
Governance context: Rapid field modification and inconsistent public naming make version-specific control arrangements difficult to establish from the current source set.
Unknowns and limits: Rapid field modification, classified software, inconsistent naming, and limited independent access make exact control arrangements difficult to verify.

Open table record

SourcesUkrainska Pravda 2023-09-04

Country record · Europe Reviewed 2026-07-31

Estonia

Manufacturer-described High for the stated functional separation

Milrem Robotics provides a clear example of separating autonomous mobility from lethal authority in the THeMIS uncrewed ground vehicle.

Public doctrine / posture: The public product example in this release separates autonomous mobility from weapon authority: navigation functions are automated while the weapon remains under human control.
Kill-chain functions: Find · Track · Target · Engage
Human control: Human-in-the-loop
Milrem states that autonomous functions are strictly limited to mobility and that the weapon system is controlled by a human operator.
Lifecycle status: Demonstrated or tested
Governance context: The functional separation is clear in the cited manufacturer statement, but customer-specific configurations and doctrine require separate verification.
Unknowns and limits: Configurations and operator doctrine vary by customer; manufacturer statements do not independently verify every deployed use.

Open table record

SourcesMilrem Robotics 2022-06-14

Cross-national supplier record · Europe / North America Reviewed 2026-07-31

Germany / Canada

Manufacturer-described High for the published control policy in demonstrated configurations

Rheinmetall Canada’s Mission Master family demonstrates autonomous mobility, networked sensing, and automated cueing while reserving kinetic decisions to humans.

Public doctrine / posture: The Mission Master example combines autonomous mobility, networked sensing, and cueing while the cited manufacturer statement reserves kinetic decisions to a human.
Kill-chain functions: Find · Track · Target · Engage
Human control: Human-in-the-loop
Rheinmetall states that targets are never engaged automatically and that a human in the loop is required for all kinetic decisions.
Lifecycle status: Demonstrated or tested
Governance context: This record describes a supplier and demonstrated configuration rather than one shared national policy or a claim about every customer deployment.
Unknowns and limits: Future configurations, customer-specific integration, and fielding scale vary and must be checked individually.

Open table record

SourcesRheinmetall 2021-08-10

Alliance framework · Transatlantic alliance Reviewed 2026-07-31

NATO (alliance framework)

Officially documented High for governance principles; NATO is not a single national operator

NATO’s public role is best understood as standard setting, interoperability, assurance, and governance across allied AI applications rather than one monolithic “NATO kill chain.”

Public doctrine / posture: NATO publishes responsible-use principles and alliance-level strategy concerning lawfulness, responsibility and accountability, explainability and traceability, reliability, governability, and bias mitigation.
Kill-chain functions: Find · Fix · Track · Target · Engage · Assess
Human control: Governance framework; no single operating mode
NATO lists lawfulness, responsibility and accountability, explainability and traceability, reliability, governability, and bias mitigation.
Lifecycle status: Policy, doctrine, or governance record
Governance context: NATO is an alliance framework rather than a single national operator. Member-state systems, rules, and operational authority vary.
Unknowns and limits: National systems, rules, and operational authority remain with allies and vary substantially.

Open table record

SourcesNATO 2024-07-10

Accessible equivalent

Country and alliance records

9 visible

Evidence Atlas country and alliance records synchronized with the map filters.
Entity	Primary evidence	Functions	Human control	Status	Key unknown
North America	Official	Find · Fix · Track · Target · Engage · Assess	Mixed or system-specific	Mixed lifecycle status across programs or configurations	Classified rules, exact model architectures, operational thresholds, and weapon-release modes for many systems are not public.
Middle East	Manufacturer	Find · Fix · Track · Target · Engage · Assess	Mixed or system-specific	Mixed lifecycle status across programs or configurations	Operational software, thresholds, deployment rules, review procedures, and many conflict-specific claims are not fully public or independently verifiable.
Europe	Manufacturer	Find · Fix · Track · Engage	Bounded post-activation autonomy	Fielded or operationally introduced	Exact software, classifier performance, abort behavior, and country-specific rules of engagement are not public.
East Asia	Corroborated	Find · Fix · Track · Target · Engage · Assess	Publicly unspecified	Developmental or research program	The maturity, combat validation, control modes, and fielded scale of particular AI-enabled kill-chain functions are often unknown.
Europe / Eurasia	Disputed	Find · Fix · Track · Engage	Publicly unspecified	Mixed lifecycle status across programs or configurations	Combat performance, error rates, software versions, operator oversight, and autonomy modes are contested or undisclosed.
Europe	Reported	Find · Fix · Track · Engage · Assess	Publicly unspecified	Mixed lifecycle status across programs or configurations	Rapid field modification, classified software, inconsistent naming, and limited independent access make exact control arrangements difficult to verify.
Europe	Manufacturer	Find · Track · Target · Engage	Human-in-the-loop	Demonstrated or tested	Configurations and operator doctrine vary by customer; manufacturer statements do not independently verify every deployed use.
Europe / North America	Manufacturer	Find · Track · Target · Engage	Human-in-the-loop	Demonstrated or tested	Future configurations, customer-specific integration, and fielding scale vary and must be checked individually.
Transatlantic alliance	Official	Find · Fix · Track · Target · Engage · Assess	Governance framework; no single operating mode	Policy, doctrine, or governance record	National systems, rules, and operational authority remain with allies and vary substantially.

Evidence is part of the visual syntax

Thirteen states prevent false equivalence.

Solid, dashed, dotted, hollow, split, and hazard-patterned records communicate source quality even before the user opens a source. Color is supplementary rather than the only signal.

Officially documented

Published policy, doctrine, program material, or another authenticated official record.

Solid border, opaque node, and heavy connection line.

Court- or treaty-recorded

A court record, treaty body, formal multilateral record, or comparable institutional document.

Double border and square marker.

Independently corroborated

The claim is supported by multiple independent public sources or a transparent verification process.

Solid border with a diamond marker.

Operator-described

A public description attributed to an operator or service user; useful but not automatically independently corroborated.

Solid border with a triangular marker.

Manufacturer-described

A supplier describes a capability, control mode, or product status. This is not independent verification of field performance.

Dashed border, reduced opacity, and an explicit claim label.

Credibly reported

Public reporting supports the claim, but the underlying primary record or complete operating context is unavailable.

Dash-dot border and circular marker.

Demonstrated in testing

A controlled demonstration or test is documented; this does not establish routine operational fielding.

Split border and laboratory marker.

Alleged

A claim exists but has not been adequately verified through reliable public evidence.

Dotted border and hollow node.

Disputed

Reliable public sources materially conflict, or a claim is contested in a way the current evidence cannot resolve.

Hazard pattern and split marker.

Operational status unspecified

Public material does not establish whether the capability is developmental, demonstrated, fielded, or routinely used.

Hollow node with an open lower edge.

Control mode unspecified

Public descriptions do not establish who authorizes force, supervises the action, or can abort it.

Blurred orbital ring and dotted border.

AI use not publicly established

Automation is documented, but no reliable public source establishes that the relevant operational function uses machine learning.

Wireframe node with an “AI?” marker.

Outdated or superseded

The claim or record has been replaced, corrected, retired, or preserved only for historical context.

Struck-through label and faded node.

System and program graph

Place each capability at the function it actually performs.

A platform can appear in several F2T2EA functions. That does not mean the software holds the same authority at every point, or that one public claim establishes the complete end-to-end chain.

Find

Detect a potential object, event, anomaly, or signal.

Fix

Establish identity, location, classification, and sufficient confidence.

Track

Maintain a time-dependent picture of movement, state, or behavior.

Target

Validate, prioritize, select a course of action, and apply governing constraints.

Engage

Apply an authorized effect or execute a bounded defensive action.

Assess

Evaluate effects, uncertainty, residual risk, and whether further action is justified.

01Official

Maven Smart System

Program / decision-support system

Official material describes sensor-data fusion, object detection, tracking, and decision support. The software is an upstream information and command layer rather than a weapon.

Find · Fix · Track · Target

02Official

Replicator portfolio

Acquisition initiative

Official program material describes accelerated fielding of attritable autonomous systems. The portfolio contains different missions, platforms, and authority arrangements rather than one common autonomy mode.

Find · Track · Target · Engage · Assess

03Manufacturer

Naval Strike Missile

Missile system

Manufacturer material describes imaging-infrared sensing and autonomous target recognition supporting terminal discrimination after a human-defined mission.

Find · Fix · Track · Engage

04Manufacturer

SPICE 250

Precision weapon system

Manufacturer material describes automatic target acquisition and recognition, autonomous functions, data-link updates, human-in-the-loop operation, and mission abort.

Fix · Track · Engage

05Manufacturer

HARPY

Anti-radiation loitering munition

The manufacturer describes autonomous search for, identification of, and attack on emitting radar targets inside a defined mission profile.

Find · Fix · Track · Engage

06Manufacturer

HAROP

Electro-optical loitering munition

Manufacturer material describes detection, recognition, acquisition, and attack; related public material distinguishes it from HARPY through man-in-the-loop employment.

Find · Fix · Track · Engage

07Manufacturer

THeMIS Combat UGV

Uncrewed ground vehicle

The manufacturer states that autonomous functions are limited to mobility while the weapon system remains under human control.

Find · Track · Target · Engage

08Manufacturer

Mission Master SP

Uncrewed ground vehicle family

The cited manufacturer demonstration combines autonomous navigation and networked cueing while stating that a human controls all kinetic decisions.

Find · Track · Target · Engage

09Disputed

Lancet-family “intelligent targeting” claim

Loitering-munition capability claim

A manufacturer and belligerent source claims intelligent targeting and recognition, while independent analysis describes a broader autonomy-development effort. The public record does not resolve control mode or performance.

Find · Fix · Track · Engage

10Reported

Saker Scout

Uncrewed aerial system

Public reporting describes AI-enabled detection and identification functions. The exact software, version, operator role, and engagement arrangement are not established by the current source set.

Find · Fix · Track · Engage

11Corroborated

PLA intelligentization and kill-web research

Doctrine and research record

Independent research supports the strategic direction toward intelligentization, decision advantage, swarms, and kill-web concepts more strongly than it supports any uniform claim about fielded capability.

Find · Fix · Track · Target · Engage · Assess

12Official

NATO responsible AI framework

Alliance policy record

Official alliance material identifies lawfulness, responsibility and accountability, explainability and traceability, reliability, governability, and bias mitigation as public principles.

Find · Fix · Track · Target · Engage · Assess

Program / decision-support systemReviewed 2026-07-31

Maven Smart System

Officially documentedU.S. Department of Defense ecosystem

Official material describes sensor-data fusion, object detection, tracking, and decision support. The software is an upstream information and command layer rather than a weapon.

Category: AI-assisted intelligence and decision support
Functions: Find · Fix · Track · Target
Documented function: Fuse heterogeneous data, surface objects and tracks, and support human analysis and command decisions.
Human-control assessment: AI assistance; no public engagement authority
No public claim in the cited source establishes independent authority to use force.
Lifecycle status: Fielded or operationally introduced
Public limitations: Operational models, thresholds, interfaces, deployments, and downstream decision rules remain partly classified or otherwise nonpublic.

SourcesU.S. Chief Digital and Artificial Intelligence Office 2026-07-23

Acquisition initiativeReviewed 2026-07-31

Replicator portfolio

Officially documentedU.S. Department of Defense / Defense Innovation Unit

Category: Attritable autonomous systems portfolio
Functions: Find · Track · Target · Engage · Assess
Documented function: Accelerate acquisition and integration of autonomous systems at scale across mission areas.
Human-control assessment: Mixed or system-specific
Control mode and engagement authority must be established for each selected system and mission.
Lifecycle status: Developmental or research program
Public limitations: A portfolio announcement does not independently establish the fielded configuration, performance, or lethal authority of every component.

SourcesDefense Innovation Unit 2024 U.S. Department of Defense 2023-01-25

Missile systemReviewed 2026-07-31

Naval Strike Missile

Manufacturer-describedNorway and allied operators; operator configuration varies

Manufacturer material describes imaging-infrared sensing and autonomous target recognition supporting terminal discrimination after a human-defined mission.

Category: Bounded post-launch navigation and terminal recognition
Functions: Find · Fix · Track · Engage
Documented function: Passive sensing, route execution, terminal target recognition, discrimination, and aim-point behavior.
Human-control assessment: Bounded post-activation autonomy
Mission authorization occurs before launch; detailed intervention and operator-specific release logic are not fully public.
Lifecycle status: Fielded or operationally introduced
Public limitations: Classifier design, operational accuracy, open-set behavior, abort capability, and national rules are not disclosed in the cited product material.

SourcesKongsberg Defence & Aerospace Current product page

Precision weapon systemReviewed 2026-07-31

SPICE 250

Manufacturer-describedIsrael-origin system and export customers

Manufacturer material describes automatic target acquisition and recognition, autonomous functions, data-link updates, human-in-the-loop operation, and mission abort.

Category: Scene matching and terminal target recognition
Functions: Fix · Track · Engage
Documented function: Electro-optical scene matching, terminal recognition, and mission updates.
Human-control assessment: Human-in-the-loop
The cited product description presents human-in-the-loop features; the exact mode, configuration, and operator procedure must still be checked.
Lifecycle status: Fielded or operationally introduced
Public limitations: Product claims do not independently establish operational classifier performance or every export-customer control arrangement.

SourcesRafael Advanced Defense Systems Current product page

Anti-radiation loitering munitionReviewed 2026-07-31

HARPY

Manufacturer-describedIsrael-origin system; operator details vary

The manufacturer describes autonomous search for, identification of, and attack on emitting radar targets inside a defined mission profile.

Category: Bounded target-profile search after activation
Functions: Find · Fix · Track · Engage
Documented function: Search for radio-frequency emissions and match them to a bounded target profile.
Human-control assessment: Bounded post-activation autonomy
The advertised mode is out of the loop after activation for the target-profile-matching function, bounded by mission parameters.
Lifecycle status: Fielded or operationally introduced
Public limitations: The public product page is a manufacturer claim and does not disclose all operator rules, abort behavior, or independent performance evidence.

SourcesIsrael Aerospace Industries Current product page

Electro-optical loitering munitionReviewed 2026-07-31

HAROP

Manufacturer-describedIsrael-origin system and reported export users

Manufacturer material describes detection, recognition, acquisition, and attack; related public material distinguishes it from HARPY through man-in-the-loop employment.

Category: Electro-optical detection and operator-mediated engagement
Functions: Find · Fix · Track · Engage
Documented function: Electro-optical search, tracking, and mission execution with operator involvement.
Human-control assessment: Human-in-the-loop
The public control description is human-in-the-loop, but not every mode and operator procedure is disclosed.
Lifecycle status: Fielded or operationally introduced
Public limitations: Combat-performance claims and exact software or control arrangements require independent verification.

SourcesIsrael Aerospace Industries Current product page

Uncrewed ground vehicleReviewed 2026-07-31

THeMIS Combat UGV

Manufacturer-describedMilrem Robotics platform; customer configurations vary

The manufacturer states that autonomous functions are limited to mobility while the weapon system remains under human control.

Category: Autonomous mobility with separate weapon control
Functions: Find · Track · Target · Engage
Documented function: Follow-me, return-home, waypoint navigation, obstacle avoidance, and integration with mission payloads.
Human-control assessment: Human-in-the-loop
Human control is retained for the weapon in the described configuration.
Lifecycle status: Demonstrated or tested
Public limitations: Customer integrations and operational doctrine vary; a manufacturer demonstration is not universal evidence for every deployed configuration.

SourcesMilrem Robotics 2022-06-14

Uncrewed ground vehicle familyReviewed 2026-07-31

Mission Master SP

Manufacturer-describedRheinmetall Canada platform; customer configurations vary

The cited manufacturer demonstration combines autonomous navigation and networked cueing while stating that a human controls all kinetic decisions.

Category: Autonomous mobility, sensing, and cueing
Functions: Find · Track · Target · Engage
Documented function: Navigation, convoy behavior, sensor sharing, and cueing between surveillance and fire-support configurations.
Human-control assessment: Human-in-the-loop
The manufacturer states that targets are never engaged automatically in the demonstrated configuration.
Lifecycle status: Demonstrated or tested
Public limitations: Future variants, customer-specific integration, and operational fielding must be assessed separately.

SourcesRheinmetall 2021-08-10

Loitering-munition capability claimReviewed 2026-07-31

Lancet-family “intelligent targeting” claim

DisputedRussian manufacturer / belligerent claim

Category: Onboard recognition and terminal mission behavior
Functions: Find · Fix · Track · Engage
Documented function: Claimed onboard recognition and operation under degraded communications or navigation.
Human-control assessment: Publicly unspecified
Who authorizes each engagement, how abort works, and which software version is used remain publicly unspecified.
Lifecycle status: Mixed lifecycle status across programs or configurations
Public limitations: Conflict claims, software versions, performance, and operator oversight are contested or undisclosed; the Atlas therefore applies the disputed state.

SourcesZALA Aero 2026-07-23 Center for Strategic and International Studies 2026

Uncrewed aerial systemReviewed 2026-07-31

Saker Scout

Credibly reportedUkrainian armed-forces use reported publicly

Public reporting describes AI-enabled detection and identification functions. The exact software, version, operator role, and engagement arrangement are not established by the current source set.

Category: AI-assisted detection and identification
Functions: Find · Fix · Track · Engage
Documented function: Onboard or associated detection and identification intended to reduce workload and communications dependence.
Human-control assessment: Publicly unspecified
AI-enabled recognition does not by itself establish autonomous weapon release; the public control mode remains unspecified.
Lifecycle status: Mixed lifecycle status across programs or configurations
Public limitations: Rapid modification, limited primary documentation, and battlefield reporting make version-specific verification difficult.

SourcesUkrainska Pravda 2023-09-04

Doctrine and research recordReviewed 2026-07-31

PLA intelligentization and kill-web research

Independently corroboratedPeople’s Liberation Army research ecosystem

Category: Human–machine integration and distributed decision architecture
Functions: Find · Fix · Track · Target · Engage · Assess
Documented function: Research and doctrine concerning AI-supported command, distributed sensing, autonomous systems, and cognitive-domain operations.
Human-control assessment: Publicly unspecified
Publicly accessible system-level engagement authority and governance rules remain limited.
Lifecycle status: Developmental or research program
Public limitations: Doctrinal ambition, research, procurement, demonstration, and operational fielding must not be treated as equivalent.

SourcesRAND Corporation 2025 National Defense University Press 2025 CNA 2026-07

Alliance policy recordReviewed 2026-07-31

NATO responsible AI framework

Officially documentedNATO alliance framework

Official alliance material identifies lawfulness, responsibility and accountability, explainability and traceability, reliability, governability, and bias mitigation as public principles.

Category: Responsible-use principles, assurance, and interoperability
Functions: Find · Fix · Track · Target · Engage · Assess
Documented function: Provide alliance-level strategy and responsible-use expectations rather than one common operational weapon mode.
Human-control assessment: Governance framework; no single operating mode
Operational authority remains with member states and varies by system, mission, and national law.
Lifecycle status: Policy, doctrine, or governance record
Public limitations: A shared strategy does not prove equivalent implementation across all allies.

SourcesNATO 2024-07-10

Timeline and changed understanding

Announcements, demonstrations, policies, and fielding are different event types.

Move the year control to view the public record as it existed at that point. Bounded dates are preserved where a source does not support day-level precision.

Show events through 2026 11 visible events

2011

Framework publicationYearOfficial

The intrusion Cyber Kill Chain is formalized

The seven-stage intrusion model establishes the defensive premise that interrupting a necessary link can prevent an adversary from reaching its objective. The Atlas keeps this cyber meaning distinct from military F2T2EA.

SourcesLockheed Martin 2011

2021-08-10

DemonstrationExact dateManufacturer

Mission Master SP demonstration is published

The manufacturer describes autonomous navigation and cueing while retaining a human-in-the-loop requirement for kinetic decisions.

SourcesRheinmetall 2021-08-10

2022-06-14

DemonstrationExact dateManufacturer

THeMIS mobility/weapon-control separation is described

Milrem states that autonomous functions are limited to mobility and that a human controls the weapon system in the demonstrated configuration.

SourcesMilrem Robotics 2022-06-14

2023-01-25

Policy updateExact dateOfficial

DoD Directive 3000.09 is updated

The official policy addresses human judgment, realistic testing, safety, cybersecurity, interfaces, legal compliance, and review for autonomy in weapon systems.

SourcesU.S. Department of Defense 2023-01-25

2023-09-04

Public reportingExact publication dateReported

Public reporting describes Saker Scout AI functions

The report supports an AI-enabled detection and identification claim but does not establish every control mode, software version, or operating configuration.

SourcesUkrainska Pravda 2023-09-04

2024

Program implementationYearOfficial

Replicator implementation is described publicly

The official program record supports accelerated fielding of attritable autonomous systems, not a single common engagement-authority claim.

SourcesDefense Innovation Unit 2024

2024-07-10

Alliance policyExact dateOfficial

NATO publishes its revised AI strategy summary

The official summary sets public responsible-use principles while leaving implementation and operational authority to allies.

SourcesNATO 2024-07-10

2025

AI-security frameworkYearOfficial

NVIDIA publishes a distinct AI-system attack chain

Recon, Poison, Hijack, Persist, Impact, and Iterate/Pivot describe attacks against AI-powered applications. This model is included to prevent it from being confused with military F2T2EA.

SourcesNVIDIA Developer 2025

2026-07-23

Official program descriptionExact publication dateOfficial

Maven Smart System program description reviewed

Official material supports sensor fusion, object detection, tracking, and decision support while leaving many operational details nonpublic.

SourcesU.S. Chief Digital and Artificial Intelligence Office 2026-07-23

2026-07-23

Manufacturer / belligerent claimExact publication dateDisputed

A Lancet-family “intelligent targeting” claim is published

The Atlas records the claim but applies the disputed and control-unknown states because performance, software, and engagement authority are not independently resolved.

SourcesZALA Aero 2026-07-23 Center for Strategic and International Studies 2026

2026-07-31

Publication releaseExact dateOfficial

KillChains.com Evidence Atlas v1.2.0 is published

The first public Atlas release adds country-level abstraction, evidence states, system records, a historical timeline, neutral comparison, saved views, and accessible table fallbacks.

Neutral comparison

Compare up to four entities without creating a “most lethal” ranking.

Use the Add to comparison controls on country and system records. The matrix asks what functions are documented, who appears to hold authority, what evidence supports the record, and what remains unknown.

0 / 4 selectedCountry, alliance, system, program, or policy records

No entities selected yet

Add a country or system above. The comparison is encoded in the current view URL so the same evidence state can be shared.

Three first-release explainers

Test the distinctions the Atlas depends on.

These compact interactions make the function–authority distinction, human-control spectrum, and chain-versus-web architecture visible without requiring the user to accept a definition on faith.

Explainer 01

AI-enabled does not necessarily mean autonomous force.

Toggle the functions attributed to software. Engagement authority is the decisive distinction.

This is AI-enabled decision support with human authority over the external effect.

Explainer 02

Human control is a spectrum, not a checkbox.

Move the authority point

Human-in-the-loopExplicit human approval is required before the modeled external effect.

Explainer 03

A chain can become a distributed kill web.

In a linear chain, interrupting the fusion node breaks the only path to the effect.

Editorial boundary

The Atlas is a publication system, not a live operational picture.

All spatial positions are abstract. The release contains country-level or broad partnership records, not current tactical deployments, active units, exact strike locations, vulnerable infrastructure, or personal target data. System cards distinguish AI assistance, algorithmic automation, autonomous navigation, automatic target recognition, human-authorized engagement, operator-supervised engagement, bounded post-activation autonomy, and publicly unspecified authority.

Publication claims retain an evidence state, source trail, review date, and visible unknowns. Manufacturer material is never presented as independent proof of operational performance. A shared Atlas view preserves the filter state and comparison selection, but it does not transmit a user identity or behavioral profile.

Read the evidence methodology, open the narrative country summaries, review the system-category guide, or inspect the source library.