Complete Documentation
Full instructions, references, and guidance for the model card.
CHAI Framework
Applied Model Card
Use the Applied Model Card to publish clear, comparable information about your AI system—covering intended use, performance, risks, and responsible AI principles.
The submit flow is an interactive Applied Model Card. Every published solution in the registry is one you can read.
1
Overview
Submit a solution
Walk through the Applied Model Card fields in an interactive form—the same structure defined by this format.
View a solution
Open a published listing to read its Applied Model Card: intended use, performance, risks, and governance details.
The applied model card describes an AI solution focused on the application in a health use case. This AI solution will be embedded within an AI system, which includes the fully operational AI use case, including the model(s), technical infrastructure, and personnel in the workflow.
This model card supports meeting the criteria for HTI-1 for predictive DSIs defined as “…technology that supports decision-making based on algorithms or models that derive relationships from training data and then produce an output that results in prediction, classification, recommendation, evaluation, or analysis.” In addition, this model card provides transparency for all Five of CHAI’s Principles of Responsible AI (Usefulness, Fairness, Safety, Transparency, Security & Privacy).
In this current draft release we have included: the complete documentation for the model card (includes: instructions, example, resources, references), a fillable template of the model card for stress-testing, and a separate copy of the example provided in the full documentation for reference.
2
Resources
Current draft versions of the CHAI Applied Model Card.
Complete Documentation - Word Version
Editable Word version of the complete documentation.
Model Card Template and Schema (GitHub repository)
Template files and schema for building and validating model cards.
Completed Example
Aidoc's BriefCase-Triage for ICH, published in the registry as a completed Applied Model Card.
3
Example
Published registry solution
BriefCase-Triage for Intracranial Hemorrhage (ICH)
Aidoc · ICH-02-RT
This is the populated exemplar from the Applied Model Card documentation, hosted as a live registry listing. Submitting a solution produces a card like this.
4
Registry adoption
On Friday, February 28, CHAI announced its partnership with Avanade on the first-ever public registry for Health AI governance. This initiative is backed by 36 partners across health systems, vendors, and collaborators.
Centralized access
Centralized access to AI model cards
Greater transparency
Greater transparency in AI development, risks, and performance.
Open access
Open access for anyone—health systems, vendors, and the public—supporting more informed AI adoption.
View supporting organizations (36)▼
- Aidoc
- Ambience Healthcare
- American Heart Association
- Avanade
- BeeKeeperAI
- Bend Health
- Better Evidence, The Global Health Delivery Project at Harvard
- Biotale Solutions
- Booz Allen Hamilton
- Cleveland Clinic
- Complear
- Dandelion Health
- Duke Health
- Dyna AI
- Ferrum Health
- Gesund.ai
- Healthvana
- Innovaccer
- Iodine Software
- Kaiser Permanente
- Lyric AI
- Memorial Sloan Kettering
- Mercy
- Mount Sinai Health System
- Nabla
- National Health Council
- OrbDoc
- Penguin AI
- Providence
- Rush University System for Health
- Sharp HealthCare
- Stanford Medicine
- Surescripts
- ThetaRho, Inc
- UMass Memorial
- University of Texas Health System
Announced at HIMSS, this registry is set to streamline AI adoption, improve accountability, and drive meaningful partnerships across healthcare. We are committed to continuous improvement and will engage members with focused sessions on integration, workflows, and feedback.
Read more in Newsweek or Fierce Healthcare.