Methodology
ATLAS is a country-level security risk monitor.
It combines official government advisories, structured security data, and multi-source public reporting into one country-level security risk model.
Short definition
ATLAS ranks country-level security risk by stacking an official baseline, a structured pressure layer, and a bounded news acceleration layer.
ATLAS is not a literal strike-probability tool. The official-first architecture anchors the score in government baselines first, structured security pressure second, and bounded news acceleration third.
The target architecture is official-first, but the current live score is still in transition. Phase 1 adds source-family and readiness scaffolding while the published score math stays on the current stable model until government advisory and broader structured integrations are ready.
Evidence order
Official / government baseline -> Structured security / operational data -> News acceleration layer
What it is not
Not an imminent-attack probability tool
Reading rule
Score and confidence are separate
01
Government advisories anchor the baseline rather than recent headlines alone.
02
Structured security data is meant to do more work than quiet strategic sensitivity.
03
News accelerates movement and confirmation but does not create the full score from scratch.
04
Low-news countries stay in the index with valid low scores.
Latest publication: Apr 5, 2026, 16:29 UTC
Model version 2026.04.05-official-first-hotfixCurrent live engine note
Government advisories are currently the main baseline, with a bounded structural context lift and a reduced news acceleration cap while structured conflict datasets are unavailable.
Score formula
Security Risk Score = Government Advisory Baseline + Structured Security Pressure + News Acceleration Layer
ATLAS is designed around an official-first hierarchy. Government advisories create the floor, structured security data anchors live pressure, and public reporting helps the model move faster without becoming the whole score.
Visual explainer
The formula is intentionally ordered. ATLAS should read as an official and structured risk model that uses news to accelerate movement, not as a headlines-only score.
Important
The score is not a probability of imminent attack or war onset.
Current rollout
Phase 1 sets the architecture and bounds the current news stack while official and broader structured families are phased in.
Government Advisory Baseline
0-40
Official advisory floor built from comparable government warning systems.
Structured Security Pressure
0-35
Structured conflict, operational warning, and spillover pressure.
News Acceleration Layer
0-25
Bounded reporting acceleration based on agreement, persistence, and pace.
ATLAS Score
0-100
The country-level security risk score shown across the product.
Three-part breakdown
The three layers at a glance
Each layer has a clear job: official baseline, structured pressure, then bounded news acceleration.
Government Advisory Baseline
0-40
40 ptsThe standing official baseline that keeps every country in the index, even when reporting volume is low.
Structured Security Pressure
0-35
35 ptsThe live pressure layer built from structured conflict, operational warning, and spillover families.
News Acceleration Layer
0-25
25 ptsThe bounded public-reporting layer that moves the score faster when multiple families keep confirming the same development.
Plain-English explanation
What the layers mean in practice
Read the model as official floor first, operational pressure second, and reporting acceleration third.
Government Advisory Baseline = the official floor
This layer asks what comparable government advisory systems are already saying about country-level security exposure. It should create the floor before fast news movement starts.
Structured Security Pressure = what the operational picture is doing
This layer responds to conflict events, operational warning systems, and real-world spillover indicators. It is designed to do more work than quiet structural sensitivity when pressure is active.
News Acceleration Layer = how quickly public reporting reinforces the move
This layer speeds the model up when multiple reporting families keep confirming the same deterioration. It is intentionally subordinate to the government and structured layers.
Low-news country logic
Low headline volume should not erase the country
In the official-first model, a quiet country still keeps a valid low score because the baseline survives even when structured pressure and news acceleration are near zero.
Old headline-heavy reading
No recent headlines
->
No visible score
A country could look as if it vanished from the model simply because the reporting window was quiet.
Official-first reading
Fiji example
Baseline
9.0
of 40
Structured
0.0
of 35
News
0.0
of 25
Total
9.0
valid low score
The government baseline still exists even when structured pressure and news acceleration are near zero. The result is a valid low score, not score disappearance.
Score vs confidence
Score level and confidence are different readings
Score answers how much security pressure and exposure the model sees. Confidence answers how hard to lean into that reading from the current evidence.
Score = security pressure level
52.0
out of 100
Score measures the combined level of official baseline exposure, structured security pressure, and bounded news acceleration.
Confidence = interpretive strength
Medium
Confidence measures how firmly ATLAS can interpret the current reading from the available evidence stack.
A country can carry a meaningful score with only medium confidence if the official or structured layers are clear but the latest public acceleration signals remain thin or uneven.
Source families
The evidence stack is official first, then structured, then public reporting
ATLAS is designed as a source-family model. Government systems anchor the baseline, structured security families anchor live pressure, and news families accelerate movement without dominating the full score.
Government baseline design
Core advisory governments are the directly normalized baseline set. Expansion governments enter first as supplementary official context. The aggregation target is a consensus rule with outlier dampening rather than a one-government trigger.
Government advisory systems
Government Advisory Baseline
Source familyOfficial advisory systems create the baseline layer. They are split into comparable core sources and slower supplementary expansion sources.
Core advisory governments
Phase 2These are the primary directly normalized advisory inputs, designed for median or weighted-consensus aggregation.
U.S. State Department · Government of Canada · Australia Smartraveller · New Zealand SafeTravel
Expansion advisory governments
SupplementaryThese enter first as qualitative or supplementary official context before more careful normalization.
UK FCDO · France foreign travel advice · Germany foreign travel and security advice · Japan overseas safety warnings · Singapore MFA travel advisories
Normalization rule
Phase 1ATLAS prefers a consensus rule with an outlier guard over reacting to a single unusually conservative government advisory.
Median or weighted consensus · Outlier dampening · Baseline floor retention
Structured security and operational data
Structured Security Pressure
Source familyThis layer is built as a source-family framework rather than an ACLED-only patch. It is where live security pressure should be anchored.
Conflict and event datasets
Live nowStructured conflict/event data should stay central to active pressure and should expand beyond one provider.
ACLED · UCDP GED
Airspace and conflict-zone warnings
Phase 3Operational warnings add real-world restriction signals that matter even before public coverage fully catches up.
EASA conflict zone bulletins · FAA restrictions · Transport Canada Safer Skies
Maritime security and transit disruption
Phase 3Shipping and maritime warning families help ATLAS register disruption around chokepoints and exposed transit corridors.
UKMTO advisories · Additional maritime warning systems
Humanitarian and strategic spillover
Phase 3Humanitarian pressure, strategic chokepoints, alliance context, and bounded nuclear sensitivity belong here as structured context, not as the whole model.
UNHCR operational context · IAEA civil nuclear context · Alliance and chokepoint references
Public reporting and acceleration signals
News Acceleration Layer
Source familyNews is the bounded acceleration layer. It helps ATLAS move faster when multiple public families agree, but it does not define the whole score alone.
Wire services
Live nowFast, factual reporting that is useful for speed, persistence, and agreement checks.
Reuters · AP · AFP
Broadcasters and international press
Live nowReadable public context that helps clarify official statements, posture shifts, and broader regional deterioration.
BBC · DW · Al Jazeera English · France24 · NHK World
Regional and theater-specific reporting
Phase 3Regional hooks are planned for later so ATLAS can add theater-specific reporting without overfitting too early.
Regional desks · Border-state reporting · Local security coverage
Implementation strategy
Official-first rollout happens in phases
The architecture is being fixed first so the product can be explained clearly now, while source ingestion expands in controlled steps rather than one large unstable jump.
Phase 1
Architecture reset and source-family scaffolding
Reframe the product around the official-first hierarchy, keep the current reporting stack bounded inside the news layer, and prepare hooks for the broader evidence families.
Redesign the score architecture and methodology around the official-first model.
Create source-family placeholders and normalization notes for government, structured, and news families.
Keep the current Reuters, BBC, DW, and Al Jazeera pipeline as the bounded news acceleration layer.
Phase 2
Core advisory normalization and broader structured pressure
Add the first comparable official baselines and one additional structured dataset so the model is no longer dependent on one structured family and one news family.
Implement core government advisory ingestion and normalization.
Add one structured dataset beyond ACLED, preferably UCDP GED.
Surface source-family visibility more explicitly in methodology and evidence framing.
Phase 3
Operational warning expansion and calibration
Expand operational warning families, maritime families, and broader reporting families, then recalibrate the score distribution once the official-first stack is materially broader.
Add EASA, FAA, UKMTO, and related operational warning families where practical.
Expand broadcaster, wire, and regional reporting families carefully.
Recalibrate score distribution after the broader evidence stack is live.
Full methodology text
Detailed layer notes, model rules, and examples
The quick explainer above is for fast reading. The detailed notes below keep the architecture defensible for deeper review.
Rule 1
Official and government signals anchor the baseline.
Rule 2
Structured event and security data anchor active pressure.
Rule 3
News accelerates movement but does not build the whole score from scratch.
Rule 4
Current direct pressure must outweigh quiet structural tension.
Rule 5
Low-news countries remain in the index with valid low scores.
Rule 6
The score reads as security risk and security pressure, not imminent attack probability.
Government Advisory Baseline
0-40
40 points
The standing official baseline that keeps every country in the index, even when reporting volume is low.
Core advisory governments are designed for direct normalization and consensus building. Expansion governments remain supplementary until they can be normalized carefully enough to compare.
Core advisory consensus
Directly normalized advisory input from the primary comparable government sources.
Expansion advisory context
Supplementary official context added carefully after core normalization is stable.
Baseline floor and outlier guard
Consensus and floor logic that keeps low-news countries in the index and dampens one-source overreaction.
Structured Security Pressure
0-35
35 points
The live pressure layer built from structured conflict, operational warning, and spillover families.
This is where direct deterioration, airspace and maritime restrictions, conflict events, and operational spillover should outweigh quiet background context.
Conflict and event datasets
Structured conflict and event families such as ACLED now and UCDP GED next.
Operational warning systems
Airspace, conflict-zone, and maritime warnings that capture real operational restriction and disruption.
Spillover and strategic context
Humanitarian spillover, chokepoints, alliance posture, and bounded nuclear sensitivity context.
News Acceleration Layer
0-25
25 points
The bounded public-reporting layer that moves the score faster when multiple families keep confirming the same development.
News helps ATLAS detect recent movement, agreement, and persistence. It should not become the whole foundation of the score.
Wire acceleration
Fast factual reporting that helps detect rapid movement without overfitting to one newsroom.
Broadcaster and international confirmation
Global public reporting that broadens confirmation and clarifies official statements and posture shifts.
Regional agreement and persistence
Hooks for theater-specific reporting, event persistence, and sustained movement across the window.
Illustrative score card
One country, shown through the new three-layer architecture
Illustrative country score
Poland
Baseline
16.0
of 40
Structured
15.0
of 35
News
8.0
of 25
Total
39.0
of 100
In the official-first model, the country stays grounded in the official baseline while structured regional pressure does most of the live work. News helps confirm pace and persistence, but it is not the whole score.
Top drivers
The example is still explained in plain language so a reader can see which layer is doing the work and why.
Government advisory baseline
BaselineComparable advisory systems keep the country inside the model even when the headline flow is quieter than neighboring flashpoints.
Structured security pressure
StructuredRegional conflict proximity, event data, and operational posture are doing the heaviest live lifting in the score.
News acceleration
AccelerationMulti-source reporting confirms the move and helps explain why the score is shifting now rather than last week.
Important notes
Read the score as structured security context, not prediction
Not a probability
A score of 80 is not an 80 percent chance of imminent attack or war onset.
No single government should dominate
Government advisories are designed to be aggregated with a consensus or median-style rule so one unusually conservative baseline does not overrun the model.
News volume is bounded
A burst of articles can accelerate movement, but the model is intentionally designed so article volume alone does not create the entire score.
Low-news does not mean invisible
Countries should remain in the index with valid low scores unless there is a real data failure rather than a quiet news cycle.