Tests a method or assurance question. It may end without an operational product.
Verified innovation watch
AI and data projects—with the safety claims kept honest.
Explore public programmes and research from aviation authorities and organizations. Each profile separates the verified purpose from what the technology cannot establish on its own.
9verified profiles
5organizations
0invented performance claims
How to read this index
Project, programme, model, and operational approval are different things.
Creates governed data, collaboration, or analysis capability. Outputs still require expert interpretation.
Can flag a pattern or estimate. It does not independently prove cause, compliance, or safety.
Depends on the applicable authority, intended function, data, validation, human role, and operating context.
Official project index
What aviation organizations are actually doing.
Status and descriptions are based on the linked official pages. Dates are shown so readers can judge whether a source describes active work, a completed project, or a historical contribution.
01
Data4Safety
A voluntary European partnership that combines contributed aviation data and collaborative expert analysis to identify systemic risks and possible mitigations.
Safety use
Supports safety-issue identification, risk assessment, performance measurement, directed studies, blind benchmarking, and discovery of system vulnerabilities.
Verification boundary
Data4Safety is a governed collaborative programme. Its existence does not make every algorithmic observation causal, operator-specific, or suitable for operational action without expert review.
Operator flight dataSafety reportsATM traffic dataWeather dataBig-data platformData science with domain experts
02
BIGDATA — New intelligence solutions
Research to mature and validate big-data and data-science methods and tools on the Data4Safety environment for wider aviation use.
Safety use
Targets new intelligence solutions for aviation safety risk management and expansion of a common reference platform and analytical methods.
Verification boundary
This is research and validation work. A project objective or prototype should not be represented as a certified operational capability or proven safety benefit.
Data4Safety data lakeBig-data technologiesData-science solutionsMethod and tool validation
03
Machine Learning Application Approval (MLEAP)
EASA-initiated Horizon Europe research into methods supporting approval of machine-learning technology for safety-related aviation applications.
Safety use
Investigated data completeness and representativeness, model generalisation, and algorithm and model robustness using aviation use cases.
Verification boundary
MLEAP provides research findings and proposed assurance methods. It is not a blanket approval of machine learning, a certification of a product, or evidence that an ML model is safe in a new context.
Learning assuranceDataset representativenessGeneralisationRobustnessW-shaped development process
04
Aviation Safety Information Analysis and Sharing (ASIAS)
A protected government–industry initiative that combines voluntarily shared and public safety data to identify emerging hazards, monitor known risks, and evaluate mitigations.
Safety use
Supports collaborative safety teams with risk monitoring, directed studies, data fusion, metrics, and evaluation of safety enhancements.
Verification boundary
ASIAS uses governed, de-identified safety data and consensus processes. Public descriptions do not expose proprietary models or justify reproducing operator conclusions from public datasets alone.
Digital flight dataVoluntary safety reportsFAA surveillance and navigation dataPublic and manufacturer dataData-fusion analytics
05
System-Wide Safety and In-Time Aviation Safety Management
NASA research developing data solutions and proactive safety-management capabilities for increasingly complex and automated airspace operations.
Safety use
Develops and demonstrates In-Time Aviation Safety Management System services and capabilities for commercial aviation and emerging operations.
Verification boundary
NASA's work is research and demonstration. An IASMS concept or prototype is not automatically an approved airline SMS process or deployable operational product.
In-time risk monitoringIntegrated safety servicesEmerging-hazard discoveryAutomation assuranceSafety demonstrators
06
ML-Enabled Safety Analytics
Machine-learning tools developed to flag hidden anomalies across large collections of flight logs and safety reports, including unusual operational patterns.
Safety use
Illustrates movement from manual review toward proactive screening of large, heterogeneous safety datasets for patterns requiring expert attention.
Verification boundary
An anomaly is a prompt for investigation, not proof of unsafe operation, causation, non-compliance, or future accident probability.
Machine learningAnomaly detectionFlight logsSafety reportsLarge-scale screening
07
Discovery of Precursors to Safety Incidents (DPSI)
Data-mining research focused on discovering sequences of events that have a higher-than-normal association with adverse aviation outcomes.
Safety use
Explores anomaly detection and precursor discovery in large heterogeneous datasets to support proactive risk identification.
Verification boundary
A statistical precursor is an association requiring validation and domain interpretation; it is not automatically causal or transferable to another fleet or operation.
Data miningSequence discoveryAnomaly detectionHeterogeneous datasetsPrecursor analysis
08
COAST machine-learning calibration
Machine-learning models trained on historical local airport traffic and meteorological data to support calibration of final-approach spacing tools.
Safety use
Supports air navigation service providers in calibrating safe and efficient time-based separation delivery for local operating conditions.
Verification boundary
COAST is an ATM spacing application, not an airline FDM risk model. Local validation and the applicable operational assurance remain essential.
Machine learningHistorical airport trafficMeteorological dataApproach spacingLocal calibration
09
Flight Data eXchange (FDX)
An aggregated, de-identified database of processed flight-data results contributed by participating airlines for safety benchmarking.
Safety use
Enables comparison by period, phase, region, airport, and aircraft category to highlight areas of safety concern and augment an operator's own analysis.
Verification boundary
IATA describes FDX as a data and benchmarking programme, not an AI model. Benchmark differences require like-for-like scope and operational interpretation.
Processed flight-data eventsDe-identificationBenchmarkingIndustry safety indicatorsPeer comparison