A better way to protect your cardholders by preventing token provisioning fraud
Tokenization helps keep commerce safe and convenient. However, fraudulently provisioned tokens threaten the payments ecosystem, with an estimated $809M losses globally in 2025.¹ VPI uses AI to help issuers differentiate between genuine and fraudulent token requests and reduce false positives.
Visa brings deep fraud prevention expertise
Visa brings decades of experience to all aspects of fraud prevention, from token provisioning requests to authorization. VPI uses segment-level supervised machine learning to identify patterns in past token requests across device, e-commerce and card-on-file tokens to better predict the probability of token provisioning fraud.
News and resources
Learn how Visa risk solutions slow fraud and keep business moving for banks.
Follow Visa Protect on LinkedIn
Stay ahead of emerging fraud with timely updates on industry trends, best practices, expert insights, and Visa Protect solutions.


- Annual token provisioning fraud as of 2025 would be $809M (up from $450M). Visa MIS; based on Visa Product Brands, VisaNet Clearing and Settlement and Fraud Reporting as of July 25th, 2025. Includes cash, excludes collections, Russia NSPK, and India Domestic volumes.
- VisaNet, issuer reported fraud rates for mobile devices using secure element or host cloud emulation in January 2025 occurring within 7 days of token provisioning. Results are based on comparing the VPI score to other device risk scores provided by token requestors in a random sample of token requests that is statistically meaningful while also maintaining Visa confidentiality obligations to other parties. The fraud prevention rate and false positive ratios comparison is determined by comparing the highest risk-rated token requestor scores versus the highest risk-rated VPI scores.
- VisaNet, Visa global token transactions starting in 2014 and as of June 2024.
- 30 years of AI and counting, September 14, 2023