Domain-agnostic intelligence infrastructure for contested knowledge domains. Built to detect the structural changes that point-in-time assessment systematically misses.
Across every contested knowledge domain — from cybersecurity to interpersonal harm to geopolitical influence — existing tools are built around events. Something happens, you assess it, you respond.
But the most consequential patterns don't announce themselves as events. They develop gradually. The underlying structure stays consistent while the visible behaviour keeps changing — adapting just enough to look different each time.
Current analytical tools are optimised for the wrong unit of analysis. They see snapshots. They miss the structural shift happening underneath.
"The same actor, operating across the same targets, using the same underlying strategy, will present completely differently to any individual assessment. That's not a bug in detection — it's a feature of how these behaviours work."
Four architectural principles govern every deployment. These aren't configuration options — they're structural properties of the methodology.
Every data source is assessed through multiple independent perspectives simultaneously. The lenses don't communicate during processing. When independent analyses converge on compatible conclusions, the resulting signal carries substantially higher confidence — because convergence was detected, not manufactured.
The engine detects change across time, not anomalies at a point. It identifies the gradual structural shifts that per-event assessment systematically misses — the patterns that are invisible in any individual snapshot but unmistakable when you see the full picture.
Every deployment is governed by constitutional constraints — hardcoded architectural invariants that cannot be overridden by users, administrators, or system updates. In domains where analytical outputs affect human safety, governance isn't optional. It's the feature.
Standard intelligence methodology: collect, process, analyse, produce. Raw data enters. Structured, provenanced intelligence emerges. Every analytical step is logged, traceable, and auditable — because intelligence without a custody chain is just opinion.
The engine is configured for a specific domain by providing a domain vocabulary, a constitutional framework, and behavioural indicators. The analytical infrastructure remains constant.
The engine identifies patterns and structural change. It does not predict what any individual will do. This is a constitutional constraint, not a limitation — prediction in human domains causes harm.
The engine processes analytical sources — research, case data, structured intelligence. It's not a monitoring tool. It analyses information that already exists; it doesn't generate collection against individuals.
Full analytical provenance. Every conclusion traces back through the fusion layer, through independent assessments, through signal validation, to the original source. Auditability is architectural, not aspirational.
The engine is domain-agnostic by design. It works wherever behavioural patterns persist across time, surface indicators change, and structured analytical infrastructure is either absent or inadequate.
The first deployment. Coercive control, technology-facilitated abuse, and domestic violence — analysed through intelligence methodology. Behavioural taxonomy, indicator library, constitutional AI governance.
Behavioural change detection for insider threat programs. The same structural persistence that characterises coercive control — adaptive surface behaviour, consistent underlying pattern — defines insider threat progression.
Longitudinal detection of coordinated influence campaigns. Identifying structural signatures that persist across platform migration, narrative pivots, and tactical adaptation.
Pattern detection across radicalisation trajectories. Identifying the structural shift from engagement through commitment that point-in-time assessment misses.
Behavioural pattern analysis for complex fraud. Detecting adaptive indicators that characterise sophisticated financial crime campaigns across evolving regulatory environments.
If your organisation works in a contested knowledge domain where patterns persist, behaviours adapt, and current tools rely on per-event assessment — we should talk.
SAFE is the first deployment — intelligence infrastructure for interpersonal harm, built on this methodology.
If you are in immediate danger, call 000. For support: 1800RESPECT · Lifeline 13 11 14. Your safety matters — reach out for help.