CMS Launches AI-Powered Fraud Detection Challenge to Safeguard Medicare Integrity

The Centers for Medicare & Medicaid Services (CMS) has unveiled the “Crushing Fraud Chili Cook-Off Competition,” a market-driven research initiative designed to leverage explainable artificial intelligence (AI) to identify fraud indicators and propose scalable solutions within the Medicare Fee-for-Service (FFS) program. The challenge invites innovative proposals that integrate machine learning (ML) with transparency and accountability, aiming to modernize fraud detection while preserving human oversight.

A Strategic Response to a Persistent Problem

Medicare’s size and complexity make it a lucrative target for fraud schemes such as false billing, upcoding, and services not rendered. These abuses erode public trust and siphon critical funds away from patient care. According to CMS, over 80% of Medicare trials or efforts to detect fraud are delayed by administrative burden and inefficient data review processes. The new competition seeks to reverse that trend by moving from reactive enforcement to proactive prevention.

The competition comes amid broader federal interest in explainable AI, techniques that prioritize not just predictive accuracy, but interpretability. For CMS, that means surfacing anomalies and patterns in Medicare claims data while ensuring regulatory teams and enforcement officers can understand and act on the system’s outputs.

Challenge Structure: Two Phases to Drive Innovation

The competition will roll out in two phases:

Phase 1 – Proposed Technology Development:
Participants submit research proposals that outline innovative uses of explainable AI tailored to Medicare FFS claims. CMS will review entries and select 10 finalists to move to Phase 2.

Phase 2 – Data Application and Modeling:
Finalists will receive access to CMS Limited Data Sets (LDS) through a secured Data Use Agreement (DUA). They will analyze 2022–2024 Standard Analytical Files covering Hospice, Part B, and Durable Medical Equipment (DME) claims for a 5% random sample of beneficiaries. Finalists must apply their models, demonstrate interpretability, and publicly share their findings.

Submissions will be judged on:

• Novelty and scalability of AI/ML techniques
• Clarity and utility of explainable outputs
• Ability to reduce manual workload while preserving human-in-the-loop integrity
• Potential to uncover systemic patterns rather than isolated fraud incidents

Explainability as a Design Imperative

Unlike traditional black-box AI models, explainable AI provides justifications for its predictions, critical in legal and regulatory contexts where due process and transparency are essential. CMS emphasizes that fraud detection models must reveal the underlying features and logic leading to anomaly identification. These features not only support oversight teams in validating findings but also help identify broader vulnerabilities across the claims ecosystem.

CMS is also clear that these solutions should transcend individual bad actors. Instead, the goal is to build indicators that flag patterns across claims and providers, enabling smarter resource allocation and broader system reform.

Data Use and Public Accountability

CMS will provide data at no cost to finalists deemed qualified under the Health Insurance Portability and Accountability Act (HIPAA) research definition. All participants must comply with CMS Data Use Agreement requirements, which include strict provisions for data storage and dissemination.

In return, CMS requires participants to publicly disseminate their findings, supporting knowledge transfer and transparency. Dissemination options may include public presentations, academic publication, or posting on affiliated websites.

Key Dates

• Phase 1 Submission Deadline: September 19, 2025
• Phase 2 Launch: October 31, 2025
• Final Submissions Due: December 1, 2025

The challenge presents a rare opportunity for developers, data scientists, and health policy experts to shape the future of fraud prevention in Medicare. By harnessing the strengths of explainable AI and structured oversight, CMS hopes to safeguard taxpayer resources while accelerating innovation across one of the nation’s most vital public programs.

More details and registration are available at the official Challenge.gov page.