The challenge

Background and context

When data from different sensors—unmanned aerial systems (UAS), distributed acoustic sensors, SIGINT (signal intelligence) receivers, EO/IR (electro-optics/infra-red) platforms, and radar—and from multiple classification domains are combined, a single error can result in compromised sources, methods, or operations. Today, compliance is enforced through manual reviews and procedural checklists that cannot keep pace with the volume and velocity of modern AI-enabled data fusion.

DND/CAF requires an automated compliance layer that sits between raw sensor ingestion and the fusion analytics pipeline. This layer must act as a policy-aware gatekeeper - tagging, filtering, and routing data according to classification markings, and operational release authorities—all without introducing latency that degrades tactical decision-making.

This challenge directly supports the CAF Digital Campaign Plan, the DND/CAF AI Strategy, NORAD (North American Aerospace Defence Command) modernization priorities, and the Cyber Forces mandate. A successful FCE becomes a reusable building block for every future fusion system—from Arctic surveillance to coalition interoperability hubs.

While compliance and data-tagging technologies exist in allied nations, Canada currently lacks a sovereign, Canadian-developed and Canadian-controlled compliance engine for multi-domain fusion. This challenge seeks to build Canadian industrial capacity and intellectual property in a critical enabling technology.

DND/CAF are hosting this challenge to observe advancements in AI technology to resolve this issue.

Examples of application can be described as, but not limited to:

Joint ISR Fusion: Automated classification enforcement when combining SIGINT, EO/IR imagery, and radar tracks. The FCE tags every data element with provenance metadata, blocks unauthorized cross-domain merges and generates an audit trail.

Maritime Domain Awareness: Compliance checks during multi-sensor anomaly detection.

Tactical Edge Dismounted: Modular compliance layer deployed on portable computer (e.g., ruggedized laptop or edge server) that enforces classification separation for wearable sensor networks operating in denied or austere environments.

DND/CAF will not provide any data (classified, unclassified, operational, synthetic, or representative) for use in training, fine-tuning, or validating AI models. Participation in this challenge assumes that innovators possess sufficient sensor-domain expertise to independently generate or obtain appropriate datasets for model development and testing.

Essential outcomes

Proposed solutions must:

• Develop a modular AI-enabled component that automatically enforces classification rules and policy constraints during multi-sensor (at least two) data fusion operations
• Apply enforcement controls based on machine-readable policy definitions across:
• Multiple sensor modalities (at least two)
• Security domains (at least Network security)
• Classification levels (at least Protected B level)
• Execute compliance checks and enforcement actions programmatically during data ingestion and fusion processing, without requiring human approval for predefined policy conditions
• Generate and retain provenance records for all data ingested into and produced by the fusion pipeline including source sensor identification, classification markings, timestamps, and domain of origin
• Produce audit logs documenting policy rules applied during fusion processing, enforcement actions taken (e.g., permit, restrict, downgrade, segregate) and resulting compliance dispositions
• Produce audit records that support traceability of data lineage from original ingestion through fusion output and are exportable for compliance review, forensic analysis, or accreditation activities

Desired outcomes

Proposed solutions should include capabilities and considerations such as, but not limited to, the following:

• Real-time compliance enforcement across multiple sensor modalities (at least two) and classification levels, with performance suitable for tactical decision-making
• Adaptable policy framework that allows compliance rules (e.g. classification guides, release authorities, coalition-specific caveats) to be updated or reconfigured without system restart, supporting rapid transition between operational contexts
• Incorporate Size/Weight/Power (SWaP) and compute limits into fusion pipelines for edge deployment and capable of maintaining compliance enforcement
• Explainability and operator trust mechanisms, including human-readable compliance decisions and controlled override capabilities with appropriate accountability safeguards