# What is a model card in AI? > A model card is a vendor's document describing an AI model: what it is for, how it was tested, known limits and risks. Some labs call it a system card. Source: https://learn.ynteractive.com/content/glossary/model-card · Updated 2026-10-20 · Free from Stepthrough (https://learn.ynteractive.com) [AI basics](https://learn.ynteractive.com/content/glossary#topic-ai-basics) · [AI glossary](https://learn.ynteractive.com/content/glossary) Quick answer A model card is a document published with an AI model that explains what it is designed for, how it was trained and tested, how it performs, and its known limits and risks. The idea comes from a 2019 research paper; today labs such as Anthropic and OpenAI publish model or system cards with major releases. Last updated October 20, 2026 [Take the free module →](https://learn.ynteractive.com/courses/ai-at-work/5)Module 5 of the free [AI at Work: Start Here](https://learn.ynteractive.com/courses/ai-at-work) course, about 10 minutes. No account needed. Updated October 20, 2026 · By [Robert Breen](https://learn.ynteractive.com/content/about) ## Why it matters for a small business When you choose an AI tool for client work, the model card is one of the few official sources on what the model is good and bad at, including safety tests and known failure modes. It is technical, but the limitations section is worth reading. Model cards don't cover the app around the model: data retention, training opt-outs and admin controls live in the product's privacy and business terms. A vendor review needs both. Model and system cards are written for a technical audience, so you don't need to read them cover to cover. Skim for the intended uses, the known limits and any safety findings that touch your kind of work. ## In a real lesson: Roll Out a Team AI Policy in One Afternoon using Claude In [Roll Out a Team AI Policy in One Afternoon](https://learn.ynteractive.com/content/ai-acceptable-use-policy-claude), Northfield Consulting Group decides that only ChatGPT Business and Claude Team on company accounts are approved. The lesson is about writing the rules, not choosing the tools. Before a tool goes on an approved list like that, the model and system cards from OpenAI and Anthropic tell you what each model was tested for and where it struggles, while the business plan terms tell you how your data is handled. Use both, and keep the policy itself to one page. [Try this lesson free](https://learn.ynteractive.com/modules/claude-team-ai-policy) or [read the step-by-step guide](https://learn.ynteractive.com/content/ai-acceptable-use-policy-claude). ## Common confusions ### Model card vs privacy policy A model card describes the model's abilities and risks. A privacy policy or data processing agreement describes what happens to your data. ### Model card vs benchmark results Benchmark scores are one part of a model card. The card also covers intended use, safety tests and limits. ## Tips - Read the intended use and limitations sections first. - Pair the card with the product's data and retention terms. - Note the model version; cards change with each release. ## Related terms [AI vendor risk assessment](https://learn.ynteractive.com/content/glossary/ai-vendor-risk-assessment) · [Approved AI tools list](https://learn.ynteractive.com/content/glossary/approved-ai-tools-list) · [AI benchmark](https://learn.ynteractive.com/content/glossary/ai-benchmark) · [Data processing agreement (DPA)](https://learn.ynteractive.com/content/glossary/data-processing-agreement) · [AI governance](https://learn.ynteractive.com/content/glossary/ai-governance) ## More AI basics terms [Model distillation](https://learn.ynteractive.com/content/glossary/model-distillation) · [Model drift](https://learn.ynteractive.com/content/glossary/model-drift) · [Model family](https://learn.ynteractive.com/content/glossary/model-family) · [Model parameters](https://learn.ynteractive.com/content/glossary/model-parameters) ## Where you use it: free lessons - [Roll Out a Team AI Policy in One Afternoon using Claude](https://learn.ynteractive.com/content/ai-acceptable-use-policy-claude) (Claude, 10 min) - [ChatGPT: Keep Client Data Safe, Set Up Your Team's Data Controls in 10 Minutes](https://learn.ynteractive.com/content/chatgpt-data-controls-protect-client-data) (ChatGPT, 10 min) ## Frequently asked questions **Where do I find a model card?** On the AI company's website, usually linked from the model's announcement or documentation. Anthropic and OpenAI call theirs system cards for some releases. **Do I need to read model cards to use AI?** Not for everyday use. They are worth a look when you approve a tool for client or sensitive work. **What is the difference between a model card and a system card?** The names overlap. A system card often covers the model plus the safety measures and product around it; a model card focuses on the model. [All AI glossary terms, A to Z](https://learn.ynteractive.com/content/glossary) · [Free prompt templates](https://learn.ynteractive.com/content/prompts)