Threat and extremism
A capability inside Nexus Prism, not a separate product. Available on a seat, or as a data feed in STIX 2.1 over TAXII 2.1.
What it correlates
Incidents, with location, actor attribution where it exists, and the method used — joined to everything else in the graph rather than sitting in its own table.
Wanted persons, cross-referenced against property, corporate and on-chain records, so a name that appears in two unrelated places surfaces as one finding instead of two.
Radicalisation indicators, drawn from the dark-web and open-source collection rather than inferred from who somebody is.
Foreign influence, tracking registered and unregistered foreign-agent activity, funding paths and the corporate structures between them.
MITRE ATT&CK mapping on the cyber side, so a technique observed here lands in your existing detection vocabulary rather than ours.
The doctrine, stated plainly
This is the pillar where getting it wrong is worst, so the rule is on the page rather than in a policy PDF:
No inferred protected-class attribute is used to select, score, rank, filter or target anyone. Not ethnicity, not religion, not national origin, and not a proxy for any of them.
We know exactly what that rule costs and exactly what it prevents, because we enforced it against our own product. In September 2026 we deleted 4,059,250 rows of surname-inferred ethnicity, withdrew 13,801 scores that had already been served to subscribers through a published feed, and removed a religion proxy that was contributing points to a threat model. The full account is on the why-different page.
Scoring is on conduct. What was done, by whom, to what, and when. Everything that survived that review is a fact about an action.
What you get
A graph you can query in one place, an alert that cites the rule that fired, and a brief that cites the rows it came from — because a finding an analyst cannot trace is a finding they cannot defend.
Published error rates. Every detector states how often it is wrong.
What this is not
Not prediction of individuals. We do not score people for what they might do. Geographic and infrastructure risk modelling is a different thing from individual threat scoring, and we do the first.
Not a watchlist product. We correlate against published government lists; we do not maintain our own list of people and we will not build you one.
Not investigation for hire. Research and analysis. Texas Occupations Code §1702.101 makes offering unlicensed investigative services a violation.