AI System Card Schema

This repository contains the AI System Card Schema, a machine-readable standard for documenting Artificial Intelligence (AI) systems. The schema is designed to promote transparency, security, and accountability in AI development by providing a structured format for sharing critical information.

---

What is an AI System Card?

An AI system card is a document that provides a comprehensive overview of a complete AI system, not just the underlying model. It acts as a standardized "label" that helps developers and users understand the system's architecture, components, data sources, and most importantly, its security and safety posture.

Similar to how model cards provide details for a single AI model, system cards extend this concept to the entire AI ecosystem, including:

• Architecture and components: Details on how the system is built and what parts it uses.
• Data pedigree: Information about the data used for training and augmentation.
• Security and safety: A record of the system's security posture and a changelog of fixed issues.

---

Why Use the AI System Card Schema?

The schema is built on an ethos of transparency, similar to open source software. By adopting a standard, open format, it enables:

• Automated Policy Checks: Tools can be built to automatically generate and consume system cards.
• Informed Decision-Making: Users can make objective, fact-based decisions about using or subscribing to an AI system.
• Community Collaboration: The open format encourages the community to inspect, audit, and contribute to the security of AI systems.

This schema is a collaborative effort to establish a common, interoperable, and machine-readable standard for documenting AI systems. We welcome contributions to help shape a more transparent and secure future for AI.

---

License

This schema is released under the Creative Commons Attribution 4.0 (CC BY 4.0) license.

Note on Usage: The AI system card for "Ask Red Hat" found at https://access.redhat.com/ai/system-card/ask-red-hat is a practical example generated by following this schema. This demonstrates how the schema can be applied to create transparent and auditable documentation for AI systems.

---

Generate HTML system card from YAML

This repo includes a minimal toolchain to validate a YAML instance against the JSON Schema and render an HTML system card using a Jinja2 template.

Prerequisites

• Python 3.9+

Install

python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt

Example data

An example instance is provided at examples/ask-red-hat.yaml.

Validate and render

python system_card_generator.py \
  examples/ask-red-hat.yaml \
  schema/system-card.schema \
  --template templates/system_card.html.j2 \
  --output build/ask-red-hat.html

If validation succeeds, the rendered HTML will be written to build/ask-red-hat.html.

MCP HTTP Server

This repository also includes a Model Context Protocol (MCP) HTTP server that exposes system card sections as read-only resources. See the mcp_server/ directory for details.

Quick start:

cd mcp_server
pip install -r requirements.txt
python mcp_server.py ../examples/ask-red-hat.yaml ../schema/system-card.schema

Notes

• The HTML template (templates/system_card.html.j2) intentionally contains no CSS so it can be copied into other portals that apply their own styling.
• The tool uses jsonschema (Draft 2020-12) for validation and Jinja2 for rendering.