Model Card Canvas
Template for structured collection of AI model card content: governance, data, business.
PURPOSE
The Model Card Canvas helps teams systematically document AI models and capture the necessary information for responsible AI use in a structured way. It addresses the growing demand for transparency and traceability of AI systems toward internal and external stakeholders. The tool is particularly relevant for meeting regulatory requirements and ethical standards.
HOW TO USE
The template is filled out for each AI model, ideally alongside model development. The fields are collaboratively completed by data scientists, product owners, and governance leads. Information on data foundations, training process, performance metrics, limitations, and business context is systematically captured. The model card is reviewed and updated with model updates.
WHAT IT IS
The Model Card Canvas is a template for structured documentation of AI models in the areas of governance, data, and business. It is based on the model card concept by Google and extends it with organization-specific fields. The template includes sections for model description, training data, performance metrics, limitations, ethical aspects, and application context.
EXAMPLE
Example: Your data science team has trained a fraud detection model for an insurance company. With the Model Card Canvas, you systematically document the training data, known limitations (e.g., underrepresentation of certain regions), the bias assessment, and the intended use cases.