One shared page for the European discussion.

Current answer: there is not one production MojoVerify-trained model behind a single fraud probability. We previously developed our own models, then retired them when AWS performed better on our U.S. results.

AWS provides Face Liveness, face comparison, face-quality signals, and core identity-document extraction. MojoVerify provides capture, local PDF417/MRZ parsing, normalization, checks, signal orchestration, thresholds, reason codes, and audit evidence. The document-authenticity and overall verification scores are deterministic scorecards; they should not be described as calibrated probabilities.

EU direction: return to a governed hybrid only for areas where the incumbent stack lacks validated coverage. Every added model or provider should be benchmarked by document version and demographic segment before production routing.

Current answer: not automatically. Production decisions do not silently retrain themselves or change from customer volume. The system is designed to rerun approved evaluation data through revised parsers, rules, thresholds, models, or providers so accuracy can be compared and improved through a deliberate release.

EU direction: create a governed feedback loop using labeled outcomes, versioned datasets, pre-release evaluation, drift and bias monitoring, human approval, rollback, and customer-visible change control. The default should be no use of customer biometric data for MojoVerify model training unless there is a documented legal basis, purpose, notice, contract, and retention policy.

Current answer: MojoWallet is not designed to custody or transmit customer funds itself. It integrates third-party providers such as Coinflow for supported fiat and crypto transaction flows. MojoWallet provides the simpler customer experience, normalized account view, ledger presentation, and orchestration; the third parties perform the underlying custody and money movement. This provider-based model works today.

For U.S. or European diligence, the funds-flow diagram should name each provider and identify which entity receives, holds, converts, or transmits each asset. Registrations and authorizations should then be confirmed against those actual roles and jurisdictions rather than describing MojoWallet as the custodian.

Current answer: gaming and sweepstakes are part of MojoWallet’s product lineage and some of its deepest implemented workflows, but that does not establish them as a majority of current customers.

EU direction: lead with the reusable identity layer—identity proofing, age and eligibility evidence, liveness, fraud controls, and verified signing. Treat gaming as a separate, country-specific track with a locally licensed operator and counsel-approved product, promotional, and payments flows.

Current answer: the capture flow accepts U.S. and Canadian driving licences and passports. Implemented non-U.S. parsing includes Canadian AAMVA PDF417 data and standard passport MRZ formats. Passports from Sweden, the United Kingdom, and Germany have also been tested successfully and worked through the existing pipeline without country-specific changes.

Those results are a strong starting point, but not yet an “all passports” or broad EU national-ID claim. AWS documents AnalyzeID for U.S.-issued identity documents, while MojoVerify’s OCR and MRZ path can process additional passports. We will expand the country, document-version, image-condition, authenticity, fraud, and regression test matrix before making broader production coverage claims.