Data classification for AI
On This Lesson 4 topics
This lesson is planned. The outline below lists the topics it will cover. The full lesson has not been published yet.
In this lesson
classify the data that can reach, pass through, or leave an AI feature, and set a rule for each class.
Planned topics
- Data can enter a model in three places: the prompt (user input), retrieved content (documents, orders), and tool results; it can leave in outputs and logs.
- Use the company's classes (public, internal, confidential, restricted) and add AI-specific rules: what may be sent to which model, and what may be stored.
- Payment card data and credentials are never allowed; personal data only with a purpose, minimization, and masking.
- Enforce by design: filter or tokenize before the model call, not by asking the model to ignore it.
- Classify outputs too: an answer built from confidential documents is confidential.
Draft outline details
Demonstration
a table for the Larkspur assistant: class, examples (public product text, order status, customer name and address, card number), allowed to hosted model? (yes, masked, no), logging rule.
Check your understanding
Where can sensitive data enter an AI feature?
Show answer
Prompts, retrieved content, and tool results, and it can leave through outputs and logs.
Practice
Apply this lesson with the module's lab, exercises and quiz.
Marking a lesson complete is optional and you can undo it any time. Your progress stays in this browser.