Digital trust is the confidence that the person, device and data behind a banking interaction are genuine, and it now has to be earned at every sensitive step, not once at onboarding. AI makes faces, voices and documents cheap to fake, so banks are layering liveness detection, device intelligence, behavioral biometrics and shared fraud data on top of identity verification.
Key points
- The US National Institute of Standards and Technology (NIST) finalized its Digital Identity Guidelines, SP 800-63-4, in 2025, adding syncable passkeys and new controls against injection attacks and deepfakes.
- The Reserve Bank of India (RBI) authentication directions, in force from 1 April 2026, encourage issuers to add risk based checks using device, location and behavioral signals on top of two factor authentication.
- The Federal Bureau of Investigation (FBI) 2025 Internet Crime Report, published in April 2026, recorded about $20.9 billion in reported US losses and, for the first time, a dedicated AI section covering 22,364 complaints and $893 million in losses.
- Deloitte's Center for Financial Services projects that generative AI could lift US fraud losses from $12.3 billion in 2023 to $40 billion by 2027.
- In a Nextyn Consulting & Research interview in May 2026, a former public sector bank manager in India expected digital trust spending to arrive in sequence: identity infrastructure first, then fraud detection, then compliance.
Method: this article draws on a Nextyn Consulting & Research interview, 20 May 2026, with a former public sector bank manager in India who now advises on privacy and data protection, plus public regulatory and industry sources. Views attributed to the interview are one practitioner's assessment.
What a modern digital trust stack looks like
A modern digital trust stack verifies identity at onboarding, then keeps confirming it through passkeys, device intelligence, behavioral biometrics and transaction context. Risk based authentication steps up checks only when signals change. Shared fraud data across institutions fills gaps that one firm cannot see, and people review consequential decisions.
FIDO2, the FIDO Alliance's open authentication standard, and passkeys replace reusable passwords with phishing resistant credentials. The FIDO Alliance's State of Passkeys 2026 report estimates about 5 billion passkeys in active use, with 75% of 11,000 surveyed consumers having enabled passkeys on at least some accounts. NIST SP 800-63-4 also covers federated identity, where one trusted provider vouches for a user across services.
The World Economic Forum's Digital Trust framework frames the same idea at board level, around security and reliability, accountability and oversight, and inclusive, ethical and responsible use. In the Nextyn Consulting & Research interview, the former banker described the operating version: trust is established at multiple steps, with continuous monitoring of identity, red flags raised in real time and access adjusted when doubt appears.
| Layer | Traditional identity verification | Continuous digital trust stack |
|---|---|---|
| When identity is checked | Once, at onboarding or login | At onboarding, login and every sensitive action |
| What is checked | Documents, a selfie, a password or one time password (OTP) | Liveness, injection attack signals, device, behavior, transaction context |
| Credential type | Passwords and SMS OTP | Passkeys and FIDO2 authenticators |
| Data used | One institution's records | Consortium and regulator data, shared fraud signals |
| Response to doubt | Approve or reject | Step up, limit the transaction, route to human review |
| Governing standard | Know your customer (KYC) and anti money laundering (AML) rules | KYC and AML plus NIST Digital Identity Guidelines and risk based authentication rules |
How AI native fraud patterns work
AI native fraud uses generative models to produce phishing, cloned voices, deepfake video and synthetic personas quickly and cheaply. The former Indian public sector banker summarized the shift as speed, scale and sophistication: messages are drafted in seconds, sent to millions, and no longer carry the spelling errors that once gave them away.
The interview described a fraud economy with specialized roles. One group collects identities, another builds malware, another places it inside corporate networks, and money mules move the proceeds. Voice cloning and video deepfakes, used in vishing and impersonation calls, were named the most prominent vector of the 12 to 18 months before May 2026.
In early 2024, an Arup employee in Hong Kong made 15 transfers totaling HK$200 million after a video call in which the chief financial officer and colleagues were deepfakes built from public footage. Entrust's 2026 Identity Fraud Report, covering September 2024 to September 2025, found that injection attacks on identity checks rose 40% over the prior year and deepfake selfies rose 58%.
| AI native fraud pattern | Primary target | Control layer that addresses it |
|---|---|---|
| Deepfake fraud on video or voice calls | Payment approvals, help desks | Deepfake detection, out of band callback, dual approval |
| Synthetic identity fraud | New credit and deposit accounts | Synthetic identity fraud detection using consortium and public records |
| Account takeover | Existing customer and employee accounts | Passkeys, behavioral biometrics, device intelligence |
| Injected selfies and forged documents | Remote onboarding | Liveness detection with injection attack detection |
| Mule account networks | Payment rails | Graph analytics and transaction fraud detection |
A Nextyn Consulting & Research fraud risk technology landscape study mapped which vendors cover which of these control layers.
What identity verification needs to add
Identity verification now needs to confirm that a real, present person stands behind valid documents, not only that the documents are genuine and match a face. That means adding liveness detection tested against ISO/IEC 30107-3, the presentation attack detection standard, plus injection attack detection and data checks that test whether the identity exists outside the file.
In the interview's words, a KYC file can hold every required document and a face, yet the face may not exist. In the Indian context, the interview noted that Aadhaar and Permanent Account Number (PAN) documents can be obtained for people whose underlying identity is questionable, so a document led KYC check will still pass.
The Unique Identification Authority of India (UIDAI) offers Aadhaar face authentication, which adds a live check against the enrolled record. For remote account opening, the RBI's Video based Customer Identification Process (V-CIP) is the channel most exposed to deepfakes, so liveness and injection checks matter most there. The interviewee expects KYC to come under pressure first.
Gartner anticipated this in February 2024, predicting that by 2026 deepfake attacks on face biometrics would lead 30% of enterprises to stop treating identity verification and authentication as reliable in isolation. Whether that share has been reached is not yet measured publicly.
How synthetic identity fraud reaches beyond customers
Synthetic identity fraud takes two forms: duplicate identities built from a real person's stolen data, and wholly invented personas assembled from real and fabricated details. The former Indian public sector banker noted that synthetic personas now apply for remote jobs, not only for credit, which extends exposure into workforce access.
TransUnion's September 2025 research put US lender exposure to suspected synthetic identities at $3.3 billion for the year ending 2024, across auto loans, credit cards and unsecured personal loans. Synthetic identity fraud detection therefore depends on cross checking applications across lenders and public records.
Remote work and cloud migration are the structural drivers cited in the interview. When a synthetic persona is onboarded as a remote employee with access to critical systems, KYC for staff becomes as important as KYC for customers.
Banking and fintech remain the primary targets because money moves directly through them, with healthcare and education data close behind.
How banks detect deepfakes and account takeover
Banks detect deepfakes by layering liveness detection, injection attack detection and media analysis with signals a deepfake cannot easily copy: device history, typing rhythm, navigation behavior and transaction context. For account takeover fraud prevention, passkeys remove reusable passwords, while behavioral biometrics flag sessions that do not match the customer's usual pattern.
Deepfake detection on its own is a moving target, because generation models improve quickly; Gartner reported that injection attacks rose 200% in 2023. The stronger approach is to make a successful deepfake insufficient. Deepfake fraud on payment approval calls is the clearest example: a cloned voice should not authorize a high value payment without an out of band confirmation, and a new device should trigger step up checks.
Support channels need the same design. The former Indian public sector banker expects customer support teams to come under pressure as AI generated callers present convincing stories of distress. Scripts, callback rules and firm limits on in call changes reduce reliance on an agent's judgment of a voice. Security operations teams face the same AI against AI problem, covered in AI SOC at APAC banks.
Why digital trust and fraud detection in banking need shared data
Fraud detection in banking works best when signals are pooled across institutions. A single bank often sees a new applicant only once or twice, with documents AI can fabricate. Consortium and regulator level data shows whether the same identity, device or address is applying at several lenders, which no single institution can see alone.
The former Indian public sector banker argued that payment fraud detection and loan checks depend on data that sits outside any one firm, at the level of the RBI, interbank exchanges and the Indian Computer Emergency Response Team (CERT-In). The same interview estimated that perhaps two or three in a hundred ransomware incidents are reported, and suggested treating breach reporting as a confidential community contribution, similar to whistleblowing.
India is building a shared fraud data layer. Reserve Bank Innovation Hub's MuleHunter.AI, which flags likely mule accounts, had been onboarded by 29 banks as of 28 July 2026, according to a government answer in the Rajya Sabha. The same answer said seven banks had joined the Digital Payments Intelligence Platform (DPIP) pilot; on 17 September 2026, RBI Deputy Governor Shirish Chandra Murmu put the pilot at eight banks.
DPIP is operated by the Indian Digital Payment Intelligence Corporation (IDPIC), a not for profit company incorporated on 16 October 2025, and the RBI's May 2026 Annual Report says the Reserve Bank is exploring full scale implementation. Shared data of this kind underpins the AI fraud detection in banking that supervisors in India now encourage. Instant rails raise the stakes, as the real time payments in Southeast Asia analysis shows.
| Market | Shared fraud and identity infrastructure | Rule that shapes bank duties |
|---|---|---|
| India | MuleHunter.AI (29 banks, July 2026), DPIP pilot run by IDPIC (eight banks, September 2026), UIDAI Aadhaar face authentication | RBI authentication directions from 1 April 2026; Digital Personal Data Protection (DPDP) Rules phased to May 2027 |
| Singapore | Singpass national digital identity; bank and telco scam controls | Monetary Authority of Singapore (MAS) and Infocomm Media Development Authority (IMDA) Shared Responsibility Framework, in force 16 December 2024 |
| Australia | Sector codes for banks, telcos and digital platforms | Scams Prevention Framework Act 2025 (assent 20 February 2025); obligations apply once sectors are designated |
Balancing behavioral monitoring with privacy
Behavioral biometrics analyze how a person types, swipes and navigates, which makes them useful for continuous authentication and transaction fraud detection. The same data is personal data under India's DPDP Act and the EU General Data Protection Regulation (GDPR), so programs need a clear purpose, minimization and governance from the start.
The former Indian public sector banker expects friction between the chief information security officer (CISO), who wants more signals, and the data protection officer, who must limit intrusion, and argued that the two roles should remain separate. The interview also suggested extending pseudonymization and tokenization beyond personal data, so that stolen records are worth less to a fraudster.
India's DPDP Rules, notified in November 2025, phase in most obligations over 18 months to May 2027. Penalties under the Act reach ₹250 crore for failing to maintain reasonable security safeguards and ₹200 crore for failing to notify a breach. CERT-In's 2022 directions already require reporting of specified cyber incidents within six hours. What each rule changes for products and contracts is the core of payments regulation consulting.
Where digital trust spending is likely to consolidate
The interview points to a clear sequence for digital trust spending. Identity infrastructure comes first, fraud detection second, and broad compliance platforms third, because compliance budgets tend to grow after penalties or mandatory reporting take effect. Governments and regulators are expected to set the shared rails, with large platforms and banks participating.
Commercial platforms, the interviewee argued, invest where it serves their own interests, so government has to coordinate the shared layer, supported by independent industry bodies. The interview also proposed carrying the DPDP Act's data fiduciary idea into information security, with security officers and ISO 27001 auditors owing disclosure duties to regulators.
| Phase | Where budgets go | What drives the decision |
|---|---|---|
| 1. Identity infrastructure | Passkeys, liveness detection, device intelligence, digital identity wallets | Onboarding losses, account takeover, NIST and RBI authentication guidance |
| 2. Fraud detection | Behavioral biometrics, graph analytics, consortium data, mule detection | Rising payment fraud and shared regulator platforms |
| 3. Compliance | Data governance, AI risk frameworks, breach reporting workflows | DPDP penalties, CERT-In reporting, audit obligations |
Vendor maturity across these layers is a technology scouting question before it is a procurement one.
On the role of AI itself, the interviewee was direct: "AI has to fight against AI," adding: "We cannot give AI a free hand without an accompanying human handler." The same gap between AI policy and enforcement appears in the Nextyn Consulting & Research AI governance in banking report.
- 01
Move high risk users to passkeys
Start with staff and high value customers.
- 02
Add device and behavior signals
Step up checks when a session stops matching the customer.
- 03
Join shared fraud data
Connect to MuleHunter.AI, DPIP and consortium checks where available.
- 04
Route AI decisions to people
Send freezes and declines on model scores to a reviewer first.
What this means for banks, fintechs, investors and regulators
Digital trust in banking rewards institutions that verify continuously, share signals and keep people accountable for consequential decisions. Each group below has a distinct first move, and the order follows the identity, fraud detection and compliance sequence described in the interview.
- Banks: Treat bank fraud prevention as continuous authentication, the core of digital trust. Move high risk customers and staff to passkeys and route consequential AI decisions to human review.
- Fintechs and identity platforms: Buyers will ask for injection attack detection, liveness detection evidence and consortium data integration, not only document checks.
- Investors: The interview points to identity infrastructure capturing value first, then fraud analytics, then compliance tooling, with regulation setting the pace.
- Regulators: Shared platforms such as DPIP and confidential breach reporting give smaller institutions signals they cannot generate alone, which strengthens fraud risk management in banks across the system.
- CISOs and data protection officers: Agree early on what behavioral data is collected, why, and for how long.
For an identity or fraud program decision, Nextyn Consulting & Research runs interviews and benchmarking with the fraud and identity leaders who have built one.
Sources
- Federal Bureau of Investigation, Internet Crime Complaint Center, 2025 Internet Crime Report, 2026. https://www.ic3.gov/AnnualReport/Reports/2025_IC3Report.pdf
- Deloitte Center for Financial Services, Generative AI is expected to magnify the risk of deepfakes and other fraud in banking, 2024. https://www.deloitte.com/us/en/insights/industry/financial-services/deepfake-banking-fraud-risk-on-the-rise.html
- NIST, SP 800-63-4 Digital Identity Guidelines, 2025. https://pages.nist.gov/800-63-4
- FIDO Alliance, The State of Passkeys 2026: Global Consumer and Workforce Report, 2026. https://fidoalliance.org/wp-content/uploads/2026/05/The-State-of-Passkeys-Global-Consumer-and-Workforce-Report-1.pdf
- World Economic Forum, Earning Digital Trust: Decision Making for Trustworthy Technologies, 2022. https://www.weforum.org/reports/earning-digital-trust-decision-making-for-trustworthy-technologies
- Reserve Bank of India, RBI issues Directions on framework on Authentication Mechanisms for Digital Payment Transactions, 2025. https://website.rbi.org.in/as/web/rbi/-/press-releases/rbi-issues-directions-on-framework-on-authentication-mechanisms-for-digital-payment-transactions
- Free Press Journal (IANS), RBI Speeds Up IDPIC Rollout To Combat Digital Payment Frauds, Pilot Running With 8 Banks, 18 September 2026. https://www.freepressjournal.in/business/rbi-speeds-up-idpic-rollout-to-combat-digital-payment-frauds-pilot-running-with-8-banks
- Rajya Sabha, Unstarred Question No. 1032, AI for Prevention of Digital Payment Frauds, answered 28 July 2026 (reproduced by TaxGuru). https://taxguru.in/?p=1063495
- Reserve Bank of India, Annual Report 2025 to 26, 2026. https://rbidocs.rbi.org.in/rdocs/AnnualReport/PDFs/0AR29052026F5B979AF274E445ABB1593EB226906335.PDF
- Press Information Bureau, RBI Strengthens Framework on Unauthorised Electronic Banking Transactions (IDPIC incorporation), 2026. https://www.pib.gov.in/PressReleasePage.aspx?PRID=2244478
- Press Information Bureau, Digital Personal Data Protection Rules, 2025 backgrounder, 2025. https://static.pib.gov.in/WriteReadData/specificdocs/documents/2025/nov/doc20251117695301.pdf
- CERT-In, Directions under Section 70B(6) of the IT Act, 2022. https://www.cert-in.org.in/PDF/CERT-In_Directions_70B_28.04.2022.pdf
- Entrust, Deepfakes, Social Engineering and Injection Attacks on the Rise: 2026 Identity Fraud Report, 2025. https://www.entrust.com/company/newsroom/deepfakes-social-engineering-and-injection-attacks-on-the-rise
- TransUnion, TransUnion Research Highlights Power of Public Data in Uncovering $3.3B Synthetic Identity Threat, September 2025. https://newsroom.transunion.com/transunion-research-highlights-power-of-public-data-in-uncovering-33b-synthetic-identity-threat/
- Gartner, Newsroom press release on deepfakes and the reliability of identity verification by 2026, February 2024. https://www.gartner.com/en/newsroom/press-releases/2024-02-01-gartner-predicts-30-percent-of-enterprises-will-consider-identity-verification-and-authentication-solutions-unreliable-in-isolation-due-to-deepfakes-by-2026
- Monetary Authority of Singapore and IMDA, Guidelines on Shared Responsibility Framework, 2024. https://www.mas.gov.sg/regulation/guidelines/guidelines-on-shared-responsibility-framework
- Australian Government, Scams Prevention Framework Act 2025 (No. 15, 2025). https://ato.gov.au/law/view/pdf/acts/20250015.pdf
- South China Morning Post, UK multinational Arup confirmed as victim of HK$200 million deepfake scam, 2024. https://scmp.com/news/hong-kong/law-and-crime/article/3263151/uk-multinational-arup-confirmed-victim-hk200-million-deepfake-scam-used-digital-version-cfo-dupe