Codoxo’s Deepfake Detection identifies AI-generated medical records for health plans

Codoxo has announced the launch of Deepfake Detection, an AI-driven fraud detection tool now being deployed by health plans across the U.S. The solution helps identify AI-generated or manipulated medical documentation and diagnostic images submitted in support of claims before payment is made.

Healthcare fraud is already a multibillion-dollar problem, and generative AI is turning documentation fraud from a manual crime into a scalable one. Fraudsters know payers rely on progress notes, clinical narratives, and diagnostic images to validate claims, and can now fabricate convincing records in minutes. This synthetic content is difficult for rules-based systems and manual review to reliably detect.

Deepfake Detection was built to close this blind spot by bringing AI-based detection and explainable risk scoring directly into payer fraud and payment integrity workflows. The launch extends Codoxo’s Point Zero philosophy by applying earlier intervention to documentation and evidence validation, not just claim submission.

“Fraudsters are adapting faster than legacy defenses can respond, and healthcare’s documentation-heavy workflows make payers uniquely vulnerable,” said Musheer Ahmed, PhD, CEO of Codoxo. “Deepfake Detection is designed to help payers fight AI-assisted fraud with AI. By identifying synthetic or manipulated medical documentation earlier, we can strengthen payment accuracy, reduce downstream recovery costs, and protect provider relationships.”

By analyzing documentation, images, and related claim context in seconds and assigning clear risk indicators, Codoxo’s Deepfake Detection helps payment integrity teams and Special Investigation Units (SIU) surface higher-risk cases within minutes. Deepfake Detection includes advanced capabilities designed specifically for healthcare fraud workflows, including:

• Cloning and duplication detection to flag reused medical records across multiple patients, exposing systematic fraud schemes that traditional reviews miss
• Partial AI-generation detection to surface blended records that combine authentic content with AI-generated content
• Behavioral cross-referencing to connect documentation with claim history and provider behavior, surfacing inconsistencies that signal elevated fraud risk
• Continuous adaptive learning to improve detection accuracy over time as fraud tactics and AI generation techniques evolve

“As generative AI becomes more accessible, verifying the authenticity of medical documentation at scale is becoming increasingly complex,” said Kurt Spear, Vice President, Financial Investigation and Provider Review, Highmark. “Healthcare organizations need new approaches to identify synthetic or manipulated documentation earlier in the process in order to protect payment integrity and reduce downstream risk.”