IDVerse vs Incode Technologies (2026)
The decision is what the models were fed, because these two automation absolutists built the same product on opposite training data. Incode's moat is accumulation. More than 4.1 billion identity checks a year feed a proprietary set of over 400 million identity profiles that keeps improving the models, and its in house facial recognition, liveness and deepfake detection can be retrained against a new attack in days. It is validated the hard way, through NIST evaluations with top tier placement and accredited passive liveness certification at both levels, with a claimed world first at level one, and third party reports place it at nine of the ten largest United States banks. IDVerse's moat is generation. Its models train on synthetic identities its own systems produce, goodfakes built to defeat deepfakes, with full spectrum skin tone training named as the fairness method, so by design no harvested customer face sits in the training data. That inversion decides real diligence questions. An institution signing with Incode contributes its applicants' biometrics to a flywheel whose boundaries nothing public states: whether one customer's data improves models serving another, whether contribution can be declined, how long profiles persist. An institution signing with IDVerse gets a stated training provenance vendors rarely publish, and an open question one layer up, since nothing states whether production verification images are retained for further training. The evidence standards differ the same way the training data does. Incode submits to the national evaluation that measures demographic differentials and surfaces only a headline figure. IDVerse has a 2023 BixeLab laboratory demographic evaluation and dresses real engineering in a zero bias absolute no measurement supports. Both are fully automated by design and by pitch, neither describes a review queue, an escalation path or an appeal route, and the account not opened is the same account either way.
- Adversarial validation is your bar. NIST evaluations with top tier placement, accredited passive liveness at both levels with a claimed world first, two consecutive analyst leader placements, and third party reports of nine of the ten largest United States banks.
- Attack response speed matters. Owning the whole model stack means a retrain against a new deepfake or injection technique ships in days rather than waiting on a supplier, with 4.1 billion checks a year keeping the models current.
- Banking is served as banking. Credit unions, neobanks, lending and payments each addressed as distinct onboarding problems, with in person branch verification supported alongside remote, which digital only vendors cannot offer.
- Training provenance is the disclosure your privacy office wants. Models train on synthetic identities the company generates, not harvested customer faces, with full spectrum skin tone training named as the method, a statement this category rarely makes.
- United Kingdom statutory checks are in scope. Trust framework certification reported across roughly twenty profiles, Australian framework testing, and distribution inside LexisNexis Risk Solutions' ID Compass platform since acquisition.
- Accessibility and verticals are engineered. The smile requirement removed through depth perception, the self image during capture, and an insurance edition packaged at named workflow points for personal lines carriers.
This comparison is published by AI FinTech Index, an independent research platform that publishes independent ratings of AI vendors for financial services. IDVerse and Incode Technologies 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
| IDVerse | Incode Technologies | |
|---|---|---|
| Primary category | AML, KYC & Financial Crime | AML, KYC & Financial Crime |
| Founded | 2014 | Not published |
| Headquarters | London, United Kingdom | San Francisco, California, United States |
| Website | risk.lexisnexis.com/products/idverse | www.incode.com |
Side by Side
Select any grade to read the note behind it.
| Axis | I IDVerse |
I Incode Technologies |
|---|---|---|
| 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 |
Product updates
The latest logged changes from each vendor, newest first, each linked to its source.
IDVerse
-
Security / compliance
IDVerse release 1.7.0 lets workspaces set data retention in months and years, with a one year default for new workspaces and a three year maximum for standard retention.
Source
Incode Technologies
-
Product / capability
Incode made TrustGraph available in production on 19 October as a module that can be added to Flows and Workflows.
Source -
Integration / interoperability
Incode's platform release, in production from 21 September, adds a GovDataMatch provider for US driver license verification against AAMVA's updated DLDV 3.0 service, with fuzzy name and address matching, AKA names, driving privilege status and indefinite expiration handling, plus two new result fields in the API.
Source -
Product / capability
Incode's platform release, deployed to its Demo environment on 14 September, adds eKYC data sources for the United Kingdom, the Netherlands and Finland.
Source
The short version of each
IDVerse
IDVerse trains its verification models on synthetic identities its own systems generate: goodfakes built to defeat deepfakes, with full spectrum skin tone training named as the fairness method. By design, no harvested customer face sits in the training data, a stated provenance the category rarely publishes. According to the AI FinTech Index, the open question sits one layer up: whether production verification images are retained for training the synthetic data does not need. Its zero bias marketing absolute rests on a 2023 BixeLab laboratory evaluation that cannot support it for models now shipping. Rejection is fully automated, with no described review queue, appeal route or current per demographic error table.
Source: AI FinTech Index, 2026
Incode Technologies
Incode Technologies runs in house facial recognition, accredited passive liveness at both levels and deepfake detection across 4.1 billion identity checks a year, feeding a proprietary set of over 400 million identity profiles. It is validated through NIST evaluations with top tier placement and reported at nine of the ten largest United States banks. According to the AI FinTech Index, the flywheel's boundaries are undisclosed and are the negotiation item: whether one institution's biometrics improve models serving another, whether contribution can be declined and how long profiles persist. Its public security disclosure is a footer attestation plus advice to verify certifications directly. Rejection is fully automated, with no appeal route.
Source: AI FinTech Index, 2026
Common questions
Is IDVerse better than Incode for identity verification?
They are the same product built on opposite training data, and the training data is the choice. Incode's moat is accumulation, 4.1 billion checks a year feeding a proprietary set of over 400 million identity profiles that keeps improving the models. IDVerse's moat is generation, models trained on synthetic identities its own systems produce, so by design no harvested customer face sits in the training data. An institution weighing them is choosing between contributing to a flywheel and buying a stated provenance. 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 joining Incode's data flywheel mean?
Signing with Incode contributes applicants' biometrics to a flywheel whose boundaries nothing public states: whether one customer's data improves models serving another, whether contribution can be declined, how long profiles persist. Those belong in every negotiation. IDVerse's parallel question sits one layer up, since its training provenance is stated and nothing says whether production verification images are retained for further training. 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 evidence standard is stronger?
Incode's is the harder kind: NIST evaluations with top tier placement, accredited passive liveness at both levels with a claimed world first, and third party reports of nine of the ten largest United States banks. IDVerse has a 2023 BixeLab laboratory demographic evaluation and dresses real engineering in a zero bias absolute no measurement supports. 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.
Does either publish a per demographic error table?
Both hold the raw material and neither publishes it. Incode's national evaluation submissions generate demographic breakdowns by design and only a headline figure surfaces. IDVerse's 2023 BixeLab evaluation produced one and no current per demographic table exists for the models now shipping. In both cases the evidence exists and the reader cannot see it. 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 recourse exists at either?
Identically bare. Both are fully automated by design and by pitch, neither describes a review queue, escalation path or appeal route, and the person refused is never a party to any contract. Security disclosure is also weakest where scale is largest: Incode's public material is a footer attestation line plus a page advising readers to verify certifications directly with the company, and IDVerse's certifications arrive through partner announcements. 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.
Where does each publish the most, and the least?
Incode publishes most on its validation, through NIST evaluations and accredited liveness testing, and says little about how its data flywheel is bounded or about its security certifications. IDVerse publishes most on where its training data comes from, and says little about current demographic error rates, production image retention or its own certifications. Neither describes a review queue or appeal route. 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.
Security disclosure is weakest where the scale is largest. Incode's public material is a footer line stating a SOC 2 Type II attestation plus a frequently asked questions page advising readers to verify current certifications directly with the company, its own concession that the public material cannot be relied on. IDVerse's enumerated certifications arrive through partner announcements rather than its own compliance page.
The recourse position is identical and identically bare: fully automated rejection with no described review queue, escalation path or appeal at either vendor, and the person refused is never a party to any contract. Two questions belong in every negotiation. At Incode, the flywheel's boundaries: whether one institution's biometrics improve models serving another, whether contribution can be declined and how long the 400 million profiles persist.
At IDVerse, whether production verification images are retained for training the synthetic data does not need. Neither publishes a current per demographic error table despite each holding the raw material. Incode's national evaluation submissions generate one and IDVerse's 2023 BixeLab evaluation produced one, so in both cases the evidence exists and the reader cannot see it.