Microblink, a provider of AI-powered identity verification, released new findings from its Fraud Lab showing a sharp escalation in AI-generated fake IDs and deepfake documents used to commit fraud across banking, hospitality, and digital onboarding.
Microblink’s Fraud Lab, an internal research and simulation team, actively generates synthetic identity documents, reconstructs deepfake patterns, and models real fraudster behavior. This allows Microblink to detect emerging attack vectors before they appear in the market and strengthen its document-verification AI to stop advanced threats.
“Until recently, producing a realistic fake ID required specialized technical knowledge,” said Microblink CEO Hartley Thompson. “Today, anyone with an internet connection can generate a convincing deepfake document using a publicly available GenAI tool. That shift has fundamentally changed fraud risk.”
Fraud Lab research shows criminals now use GenAI to generate and alter a wide range of document elements, enabling impersonation, account takeover, and large-scale identity fraud. Examples include state and national ID templates, face-swapped IDs using stolen real identities, blended synthetic identities and fabricated documents with realistic security features.
Microblink reports that these synthetic deepfake documents are increasingly used to:
- Apply for loans or credit using stolen identities
- Open accounts or access financial services
- Check into hotels using forged documents
- Evade KYC controls across regulated industries
Unlike providers that rely primarily on large historical data graphs or external data sources to detect fraudulent behavior, Microblink uses a proactive, simulation-focused approach.
“Most systems wait for fraud to show up in customer logs,” Thomspon said. “Our models are trained on the attacks we generate ourselves, which lets us respond before those attacks spread.”
The Fraud Lab generates synthetic fraudulent IDs at scale, replicating thousands of attack variations simultaneously. This capability allows Microblink to detect subtle manipulations that many identity providers miss, including:
- Face-swap manipulations on official IDs
- AI-generated photos blended onto legitimate templates
- Fabricated documents where all security elements appear visually correct
- Injection-attack patterns mimicking real fraudster behavior
Real-World impact: 1300 fraudulent documents blocked weekly
One major fintech client serving the credit union sector is now detecting more than 1,300 fraudulent identity documents every week using Microblink Fraud Lab’s enhanced capabilities to more accurately identify documents that previously could have passed onboarding.
Fraud Lab detection metrics and trends:
- 23% increase in AI-generated fake IDs detected over the last 12 months
- Fraud Lab identified 24 synthetic unique documents across 40 geographies
- Fraud Lab improved coverage speed by 46% compared to previous manual review methods
- Face-swap document fraud increased 200% year-over-year
Preparing for 2026: Regulatory and compliance advantages
A major trend identified by the Fraud Lab is rising regulatory scrutiny around training-data transparency, including new requirements under the EU AI Act. Companies relying heavily on personal documents for model training may face increased compliance risks.
“Because we train on synthetic data instead of sensitive customer IDs, we avoid many of the regulatory and privacy constraints that competitors are now struggling with,” Thompson said. “Synthetic data expands what we can cover while strengthening compliance.”