Jumio

Tamper checks on document images, selfie to document matching and liveness against printed photos or synthesized video are all inference. Without the models, what remains is a photo upload form.

Verification runs in real time, and the privacy notice places the decision with the customer, saying the firm itself makes no automated decisions with legal effect. Review queues, escalation thresholds and a route to a person for a rejected applicant are not described.

Liveness detection reached ISO 30107 part 3 Level 2 compliance, announced in October 2025, an external presentation attack test of one component. Document authenticity accuracy, face match error rates and drift monitoring for the rest of the pipeline are unpublished.

Stanleybet reports onboarding time cut in half, a named figure from gaming rather than financial services, with no period or method. More than one billion transactions processed and over 5,000 supported ID types round out the claims, with Alaska Airlines and Personal Pay as further named case studies.

Analytics and research to improve the services are a named purpose for end user images and biometrics, and the Jumio Identity Graph of tens of millions of identities draws on activity across customers. Whether one institution's end users train models serving others, or whether an institution can opt out, is not stated.

The online services privacy notice of August 4, 2026 covers end users verified for customers, names document images, selfies and facial biometric data, and commits to destroying biometric data within three years of submission, or sooner on customer instruction. No subprocessor list is published.

A PCI DSS section sits on the compliance pages, while the trust center is closed to the public. SOC 2 and ISO 27001 status are not stated publicly.

Compliance pages map products to Swiss FINMA rules, Canadian FINTRAC rules and PSD2, along with Malaysia's eKYC guidance and data protection law, and European electronic identity accreditation is managed for customers in close to twenty countries. Behind the position stands no register entry or admission.

Face matching at account opening carries a documented risk of error rates that vary by skin tone, age and sex, and a false non match denies access to banking. Its short reference page claims training on globally distributed data and nearly equal accuracy across geographies, but gives no method, group error rates or independent evaluation.

Rights to access, change or delete data reach end users through the customer or by email to the firm, a correction route over service data. Accuracy warranties and indemnities are unpublished, and a rejected applicant is offered no published path to a second review.

Service providers are named only by category, cloud providers, fraud prevention services and government databases, and proprietary biometric technology is described without any model or hosting provider. Third party access to facial images is never addressed.

Government issued digital credentials are accepted across more than sixty countries and territories, and Brazil's digital driving license is validated against the issuing authority's records. AWS, Microsoft and Oracle appear as technology partners, and developer documentation is public.

End user data may be processed in the United States, with transfers from Europe, the UK and Brazil under adequacy decisions or standard contractual clauses. Hosting regions, residency commitments and a single tenant option are not published.

Per transaction pricing is the norm in identity verification, and neither a unit price nor volume tiers are published. Without the one unit the relationship is built on, a buyer cannot estimate cost.

Financial services is a named segment spanning banking, lending and payments with due diligence and AML framing, beside crypto, gaming and travel. Healthcare, telecommunications and government add further breadth.