Callsign vs Oz Forensics (2026)
The decision is which moment of identity you are defending, the session or the check. Callsign recognizes a returning customer continuously, inferring identity from how a person types, swipes and holds the device. It fuses device intelligence, contextual analytics and telecommunications signals that reveal a live call during a payment, and an orchestration engine decides what assurance each interaction actually needs. Oz Forensics proves a moment: that a live person is present and that two faces belong to the same person. It is sold as the component other platforms embed, with liveness running actively or passively, on device or on server, and an on premises license that keeps biometric processing inside the institution entirely. The sharpest separation is who gets told. Callsign's intervention design does something fraud tooling rarely does: on detecting social engineering it warns the person about to send the money, in the moment, converting a silent institutional risk decision into information the affected person can act on. Oz's failure mode is the inverse: a liveness rejection presents as a failed attempt with a prompt to retry, so a person repeatedly refused by a model may never learn a model was involved. The assurance postures invert the same way. Oz launched a public trust center in 2026 carrying ISO 27001, SOC 2 Type II and the subprocessors it works with, a step this category rarely takes. That sits on top of presentation attack testing at three levels by iBeta and BixeLab and a NIST face recognition evaluation submission. Callsign's assurance exists privately: tier one deployments imply repeated vendor vetting, and no attestation or trust surface is published for a buyer to read.
- Scams are your loss line. Real time social engineering detection with a contextual warning to the customer before the transfer, plus telecommunications network signals showing a live call during a payment, is built against the Payment Systems Regulator's reimbursement regime, which moved scam liability onto banks.
- You are defending journeys, not checks. Login, payments, account creation and network access run through one orchestration engine across five named fraud vectors, so assurance strengthens where risk appears rather than at a fixed gate.
- Both sides of the trade off are published. Forty percent more fraud detected while holding 99 percent acceptance states detection and customer experience together, and independent analyst documentation names the three usage dimensions that drive cost.
- Biometric data must not leave your estate. An on premises license runs liveness and matching entirely inside your environment, on device processing narrows exposure further in the hosted path, and the trust center names the subprocessors behind the hosted service.
- External validation decides your model risk review. Presentation attack detection at levels one, two and three by iBeta and BixeLab, separate injection attack assessment, and a NIST face recognition evaluation submission is testing the vendor does not control.
- You embed rather than deploy. Web and mobile kits with on device or server execution serve platforms and institutions building their own flow, and other identity verification vendors embedding these same models is a harder integration test than any direct deployment.
This comparison is published by AI FinTech Index, an independent research platform that publishes independent ratings of AI vendors for financial services. Callsign and Oz Forensics are each graded against the same capability taxonomy, from each vendor's own public materials and the regulatory record, under the AI FinTech Index verification standard. No vendor pays for placement, and no vendor has reviewed this page. How this evidence is graded
Plain facts
| Callsign | Oz Forensics | |
|---|---|---|
| Primary category | Fraud Detection & Transaction Risk | Fraud Detection & Transaction Risk |
| Founded | Not published | 2017 |
| Headquarters | London, United Kingdom | Dubai, United Arab Emirates |
| Website | www.callsign.com | ozforensics.com |
Side by Side
Select any grade to read the note behind it.
| Axis | C Callsign |
O Oz Forensics |
|---|---|---|
| AI Centrality | ||
| Autonomy and Oversight Model | ||
| Model Risk Management and Transparency | ||
| Operational and Outcome Evidence | ||
| AI Safety and Data Stewardship | ||
| GLBA and Data Privacy Posture | ||
| Security Certifications and Trust Center | ||
| Regulatory Status and Licensure | ||
| AI Governance and Bias Disclosure | ||
| AI Liability and Recourse | ||
| Model Supply Chain Disclosure | ||
| Core Systems and Integration Depth | ||
| Deployment Model and Data Residency | ||
| Commercial Transparency | ||
| Institution and Segment Coverage |
The short version of each
Callsign
Callsign recognizes a returning customer continuously, inferring identity from how a person types, swipes and holds the device. It fuses device intelligence, contextual analytics and telecommunications signals that reveal a live call during a payment, and an orchestration engine decides what assurance each interaction actually needs. According to the AI FinTech Index, its intervention design does something fraud tooling rarely does: on detecting social engineering it warns the person about to send the money, in the moment, converting a silent institutional risk decision into information the affected person can act on. The gaps sit beside the design. Behavioral profiling varies with age, motor control, disability and device, and no demographic breakdown is published. The 99.999 percent precision claim carries no methodology, sample or definition. No attestation or trust surface is published for a buyer to read, and the assurance exists privately in tier one vendor vetting.
Source: AI FinTech Index, 2026
Oz Forensics
Oz Forensics proves a moment of identity: that a live person is present and that two faces belong to the same person. It is sold as the component other platforms embed, with liveness running actively or passively, on device or on server, and an on premises license keeping biometric processing inside the institution entirely. According to the AI FinTech Index, its assurance is public. A trust center launched in 2026 carries ISO 27001, SOC 2 Type II and its subprocessors, a step this category rarely takes, on top of presentation attack testing at three levels by iBeta and BixeLab and a NIST face recognition evaluation submission. Its integration creates a failure mode. A liveness rejection presents as a failed attempt with a prompt to retry, so a person repeatedly refused by a model may never learn a model was involved. Only a single headline accuracy figure is surfaced rather than the demographic differentials its own evaluation submission generates.
Source: AI FinTech Index, 2026
Common questions
Do Callsign and Oz Forensics do the same job?
Different moments of identity. Callsign recognizes a returning customer continuously from typing, swiping and device signals during a session, while Oz Forensics proves a moment, that a live person is present and two faces match, as a component other platforms embed. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified August 23, 2026. No vendor pays for placement.
Which product tells the affected person anything?
Callsign's: on detecting social engineering it warns the person about to send money, in the moment. Oz's inverse failure mode is that a liveness rejection presents as a failed attempt to retry, so a person refused by a model may never learn one was involved. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified August 23, 2026. No vendor pays for placement.
Whose assurance is publicly readable?
Oz Forensics publishes a 2026 trust center with ISO 27001, SOC 2 Type II and its subprocessors, plus presentation attack testing at three levels by the accredited laboratories iBeta and BixeLab and a NIST face recognition evaluation submission, an artifact set this category rarely publishes. Callsign's assurance exists privately behind tier one deployments. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified August 23, 2026. No vendor pays for placement.
What does neither vendor show?
Neither publishes performance across the populations its own modality disadvantages, behavioral profiles across age and disability at Callsign, demographic differentials at Oz, and Callsign's 99.999 percent precision claim carries no methodology. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified August 23, 2026. No vendor pays for placement.
Related comparisons
Other published head to head assessments involving these vendors or their closest peers. The full set for this category is on the Fraud Detection & Transaction Risk page.
Neither vendor publishes performance across the populations its own modality is known to disadvantage: Callsign profiles typing and swiping, where age, motor control, disability and device drive variation, and publishes no breakdown, while Oz surfaces a single headline accuracy figure rather than the demographic differentials its own NIST evaluation submission generates, with no analysis across the emerging market capture conditions that form its footprint.
One published number needs its working shown, Callsign's 99.999 percent precision claim for behavioral recognition carries no methodology, sample or definition. A person wrongly refused has no described route at either vendor, though only one of the two products will even reveal to them that a decision occurred.