Overview

Document verification serves as the first line of defense in any master servicing workflow for structured finance facilities. Before a single loan can be included in a borrowing base calculation, we must confirm that the underlying documentation is authentic, consistent, and meets all facility eligibility criteria.

In 2025, this mission has become dramatically more complex. The emergence of sophisticated AI image generation tools—including Google's Gemini (Nano Banana), ChatGPT's DALL-E, Midjourney, and others—has fundamentally changed the threat landscape. Fraudsters can now generate convincing fake identity documents, bank statements, employment letters, and loan agreements in minutes, with no technical expertise required.

At Cascade, we actively monitor emerging fraud vectors, including AI-generated synthetic documents. We're continuously evaluating detection technologies and refining our verification processes to stay ahead of these evolving risks.

Key Outcomes

The Challenge: AI-Generated Document Fraud

The document verification challenge has evolved beyond simple forgery detection. Today's threat comes from generative AI tools that can produce near-perfect synthetic documents at scale.

The Scale of the Problem

Research from Sumsub's Q1 2025 Identity Fraud Report reveals alarming statistics:

• 311% increase in synthetic identity document fraud in North America (Q1 2025 vs Q1 2024) [1]

• 1,100% increase in deepfake fraud year-over-year [1]

• 50% of all identity fraud attempts now involve fake IDs and forged documents [2]

• $40 billion in projected US fraud losses by 2027, up from $12.3 billion in 2023 (Deloitte Center for Financial Services) [3]

• Over 76% of fraud and risk professionals report their organizations faced AI-driven attacks in 2024 [4]

Documents at Risk

Every document type in a loan verification package can now be synthetically generated:

‍

Real-World Case Study

Detecting AI-Generated Identity Documents

Gemini’s content filtering is highly inconsistent, changing its responses based on language, timing, or slight wording shifts. Because the system reacts to keywords instead of context, even standard business research can be incorrectly flagged or blocked. To demonstrate the severity of the AI fraud threat, and Cascade's detection capabilities, we conducted a controlled test using Google's Gemini image generation tool (Nano Banana).

The Test: Creating a Synthetic Indonesian KTP

Using simple text prompts, we generated a fake Indonesian national identity card (Kartu Tanda Penduduk or KTP) in under three minutes. The prompt requested a realistic-looking ID photographed on a surface—the type of image commonly submitted for KYC verification.

What Makes This Dangerous

At first glance, this document appears authentic. It includes:

• Correct Indonesian KTP layout and color scheme

• Province and regency headers (Provinsi Jawa Barat, Kabupaten Sukabumi)

• NIK (national identification number) field

• Photo with realistic lighting and positioning

• Official watermarks and background patterns

• Signature block and official stamp area

• Realistic surface texture suggesting physical card photographed

This is exactly the type of document that would pass manual visual inspection in a traditional KYC process.

How Cascade Detected the Fraud

Our four-layer verification system identified multiple fraud indicators:

‍

‍Key Benefits

For Lenders

• AI Fraud Protection: Multi-layer detection catches synthetic documents that would pass manual review, protecting collateral quality

• Real-Time Borrowing Base Integrity: Only verified, authentic collateral included in calculations with continuous visibility

• Complete Audit Trail: All AI detection results, cross-check outcomes, and eligibility assessments stored with timestamps

• Consistent Standards: Automated verification applies eligibility criteria uniformly, eliminating human subjectivity

For Originators

• Fraud Risk Mitigation: Identify synthetic identities before they enter your pipeline, protecting your facility relationship

• Rapid Eligibility Decisions: Processing in minutes instead of days accelerates access to borrowing capacity

• Clear Exception Management: Specific reason codes for rejected documents enable fast remediation or removal

• Predictable Outcomes: Consistent application of standards reduces disputes about borrowing base calculations

For Both

• Real-Time Dashboards: Continuous visibility into eligible vs. ineligible breakdown, AI detection flags, and remediation status

• Regulatory Preparedness: Complete documentation of AI fraud detection for auditor and regulator inquiries

• Operational Efficiency: Single-pass verification eliminates multiple rounds of document requests and manual review

• Portfolio Insights: Identify systemic documentation issues, track fraud attempt patterns, and monitor verification trends

Conclusion

The emergence of AI-generated document fraud has fundamentally changed the risk landscape for structured finance facilities. Synthetic identity documents can now be created in minutes by anyone with access to consumer AI tools. Traditional manual verification methods are no longer sufficient.

Cascade's document verification solution addresses this new reality with a four-layer approach that combines:

1. AI-generated content detection

2. Intelligent document processing with jurisdiction-aware OCR

3. Comprehensive cross-validation across all loan package documents

4. Eligibility criteria validation against facility requirements

This eliminates the historical trade-off between verification thoroughness and operational speed. Lenders gain confidence in collateral authenticity. Originators gain scalability without fraud exposure. Both gain the transparency and audit trails that modern structured finance demands.

‍

Ready to Protect Your Facility?

Contact Cascade to schedule a demonstration of our AI-powered document verification solution.

‍

Sources

All statistics cited in this white paper have been verified from the following authoritative sources:

[1] Sumsub Q1 2025 Identity Fraud Trends

"Synthetic Identity Document Fraud Surges 300% in the U.S. - Sumsub Warns E-Commerce, Healthtech and Fintech at Risk"

https://sumsub.com/newsroom/synthetic-identity-document-fraud-surges-300-in-the-u-s-sumsub-warns-e-commerce-healthtech-and-fintech-at-risk/

[2] Sumsub Identity Fraud Report 2025-2026

"Fraud Trends for 2025: From AI-Driven Scams to Identity Theft and Fraud Democratization"

https://sumsub.com/blog/fraud-trends-sumsub-fraud-report/

[3] Deloitte Center for Financial Services

"Generative AI is expected to magnify the risk of deepfakes and other fraud in banking"

https://www.deloitte.com/us/en/insights/industry/financial-services/deepfake-banking-fraud-risk-on-the-rise.html

[4] AllAboutAI - AI Fraud Detection Statistics 2025

"AI Fraud Detection Statistics 2025: 50x Faster Detection & 98% Accuracy"

https://www.allaboutai.com/resources/ai-statistics/ai-fraud-detection/

[5] Google DeepMind - SynthID

"SynthID - Tools for watermarking and identifying AI-generated content"

https://deepmind.google/models/synthid/

[6] Google Blog - AI Image Verification

"How we're bringing AI image verification to the Gemini app"

https://blog.google/technology/ai/ai-image-verification-gemini-app/

[7] Google Blog - SynthID Detector

"SynthID Detector: Identify content made with Google's AI tools"

https://blog.google/technology/ai/google-synthid-ai-content-detector/

All demonstration data is fictional. The AI-generated document shown was created solely for illustrative purposes.

‍