Extend Verify API Fights AI-Generated Document Fraud
Extend launches Verify, an API designed to help AI agents detect fraudulent documents like invoices and bank statements before they influence critical business decisions.
Extend has introduced Extend Verify, a new API aimed at bolstering the integrity of AI-driven decision-making processes by detecting document fraud. As businesses increasingly delegate tasks to AI agents, the authenticity of the documents these agents process—such as bank statements for loan approvals, invoices for payments, or receipts for insurance claims—becomes paramount.
The proliferation of sophisticated AI tools has made it easier than ever to generate convincing forged documents, posing a significant risk to organizations. When fraudulent documents bypass human review and lead to incorrect decisions, the financial repercussions can be substantial, potentially costing businesses millions. Extend Verify seeks to bridge this trust gap, enabling AI agents to flag suspicious inputs before they trigger erroneous actions.
"Businesses want to move beyond agents that search documents to agents that help get work done in the real economy," stated Kushal Byatnal, CEO of Extend. "That requires trust in the evidence those agents act on. Verify helps developers build document integrity into their pipelines to prevent fraud-related losses, mitigate the difficulty of human review, and scale up agent deployments."
The Extend Verify API employs a multi-faceted approach to document integrity assessment. It meticulously examines visual content, file structure, and revision history, integrating specialized AI models, forensic checks, and agentic analysis. This comprehensive analysis aims to identify a range of suspicious signals, including entirely AI-generated documents, localized tampering within otherwise legitimate files, conventional digital alterations made with PDF editing software, and even changes made after a document has been digitally signed.
Upon analysis, Verify returns detailed findings, including detected signals, risk assessments with accompanying scores, explanations for the identified risks, and traceable evidence. This information empowers businesses to make informed decisions about which documents warrant further scrutiny, thereby optimizing their fraud detection workflows.
In a benchmark evaluation designed to test its efficacy, Extend Verify demonstrated strong performance, accurately detecting 96.3% of fraudulent documents while maintaining a low false-positive rate of 4.0%. This balance is crucial for practical deployment, ensuring that legitimate documents are not unduly flagged, which could otherwise overwhelm review teams.
For instance, a lender could leverage Verify to scrutinize income documents before approving a loan, a payments team could use it to identify altered invoices before authorizing a transaction, and an insurer could employ it to detect fabricated receipts before settling a claim. The API can be used independently via a standard API call or integrated into existing document processing pipelines that already utilize Extend's services for parsing, splitting, and extracting data.
By consolidating document processing and integrity checks within a single platform, Extend aims to provide AI agents and applications with not only structured information but also verifiable evidence regarding the source and authenticity of the data they rely upon, thereby enhancing security and operational efficiency.