0%
ReadingGetting what you want9 min, about 660 words

What not to paste in

Choose appropriate information and tasks before sending anything to an AI service.

Make the decision before sending

Before a prompt leaves your device, decide whether the information belongs in that service. Products have different settings and data practices. Do not assume a conversation is a private notebook simply because only you can see its chat window.

For this course, invented names and small made-up datasets usually teach the same skill as real private records. Use the safer practice example from the start; deleting a message later is not a substitute for choosing carefully before sending it.

Rewrite a risky example

Suppose you want help writing an email about a schedule conflict. A draft includes a classmate's full name, phone number, home address, and a medical explanation. The job only needs: “A teammate cannot attend Tuesday. Ask whether we can meet Thursday.”

Remove unnecessary details and replace identities with roles such as teammate. Then ask for help with tone or structure. This is data minimisation in practical language: provide the information the task needs and leave out what it does not.

Do not include passwords, access tokens, private account details, or confidential documents. Information about someone else is not automatically yours to share. If a school or organisation provides an approved tool, follow its rules and ask the responsible person when you are unsure.

Use a tutor in a way that teaches you

For school work, begin with the assignment's rules. If you must write the answer independently, ask for an explanation of a concept, feedback on your attempt, or another practice example. Having a system produce the final submission may violate those rules and leave you unable to explain your own work.

A useful request is: “Here is my attempt. Point out the first step whose reasoning is incorrect, and explain the relevant idea.” After receiving help, solve a similar problem yourself. If you cannot, you have found what to review.

Keep track of assistance you need to acknowledge. Different teachers have different expectations; the course cannot replace their instructions.

Match the tool to the task

For arithmetic, use a calculator or an actual calculation tool and check the inputs. For an event that happened recently, look for current primary sources. For a claim about a document, provide and inspect the relevant document. Some AI applications can use these tools; a text answer alone does not establish that they did.

For an important decision, identify who or what can provide dependable evidence. A fluent suggestion can help formulate questions without being sufficient grounds to act.

A short pre-send check

Ask yourself: Does this contain private information? Do I have the right to share it? Does the task need every detail? What would happen if the response were wrong? Am I using help in a way that follows the rules for my work?

If you cannot answer comfortably, change the task or the information before sending it. You can learn prompting perfectly well with invented club notices, fictional timetables, and the public examples in these lessons.

Transform a realistic request without carrying private details

Imagine a club signup sheet containing student names, contact details, and snack preferences. You want help making a snack-count table. The model does not need names or contact details to explain counting. Replace the sheet with a fictional sample: Person A, apples; Person B, crackers; Person C, apples. Ask for a counting method using that sample.

Removing a name alone may not make a record anonymous. An unusual combination of school, age, activity, and schedule can still identify someone. Start from the minimum information necessary for the task instead of copying the whole document and deleting one column afterwards.

The mini editors in later lessons let you personalise examples. Use an invented name or age there. The purpose is learning how inputs affect outputs, not supplying a real identity. Local browser work can still remain visible to another person using the same browser, so choose examples you are comfortable leaving in your project.

Keep your progress

Sign in and every reading, quiz, and exercise you finish is saved.

Sign in