How AI fraud Is evolving faster than AP & procurement defenses (www.techradar.com)

Fraudsters have moved past clumsy phishing and now use advanced AI—voice cloning, deepfake video calls and synthetic invoices that replicate vendor logos, formatting, payment history and metadata—to execute targeted, multi-stage campaigns against accounts payable (AP) and procurement. The scale is already serious: organizations lose about 5% of revenue to fraud on average, the FTC logged 845,000 imposter scams in 2024, and high-profile losses (a $25M deepfake-based breach) show attackers can bypass human intuition and basic validation checks by mimicking legitimate workflows. The defense shift is technical and operational: rules-based controls and manual reviews aren’t enough. AI-powered anomaly detection trained on an organization’s own payment history can spot subtle deviations—e.g., unusual sequences of routing numbers or atypical payment behaviors—that reveal synthetic invoice rings or account-takeover attempts. Real-time cross-referencing of vendor bank-account changes against trusted databases, continuous vendor verification, deepfake awareness training, zero-trust payment protocols and cross-functional incident plans are recommended. The takeaway for AI/ML teams and finance leaders: treat fraud as an adversarial, real-time problem—deploy bespoke ML models, automate verifications, and assume attackers will match your tooling unless detection and process design stay ahead.