In research
The science behind RegulixOne.
Explainable risk decisions, deepfake-resistant onboarding and precise Arabic name matching.
The problem
Deepfakes and synthetic identities now target digital onboarding. Risk decisions made by opaque models are hard to defend to a regulator. And laundering networks hide across many accounts that look ordinary one at a time.
Questions we are researching
- How can onboarding detect face swaps, voice clones, injected video and generated identity documents?
- How can every risk score carry a plain-language reason that a regulator accepts?
- How can connected accounts, mule networks and hidden beneficial owners be found across customers?
- How can AI prepare compliance paperwork for an officer's review, with the officer always deciding?
What we plan to build
Planned outputs. Each passes the lab's stage gates before it is offered to anyone.
- Deepfake defence research
- Detection methods for onboarding, tuned for UAE documents and Arabic speech.
- Explainability layer
- A clear reason attached to every risk decision.
- Network analysis
- Tools for investigators to see relationships across customers.