What Should Never Go into an AI Prompt
The moment you send a question to AI, what you typed leaves your hands. Here's the information teachers and students must keep out of a prompt.
Wanting to polish a draft for the student record quickly, a teacher pastes a whole passage containing a student's real name and family circumstances into a chatbot. The result is satisfying, but in that moment one student's sensitive private life flowed onto an outside server. Once entered, text can't be recalled, and it's hard to know where it is stored. Everything you type into a prompt has to be information you would be fine with someone else seeing. You need a clear line around what must not go in.
The List of What Must Not Go In
The following never goes into a prompt, for any convenience whatsoever. Post this list at the edge of your desk and it can stop a mistake even in a rushed moment.
- Identifying information: A student's real name, national ID number, address, phone number, photo.
- Sensitive information: Health or disability, family circumstances, religion, records of school violence, and the like.
- Other people's information: Assessments of, or private matters about, a colleague or another student.
- Security and confidential material: Exam questions, passwords, internal school documents.
Remember above all that the student record, counseling notes, and anything related to school violence are by their nature highly sensitive information.
Safe Workarounds
You can get plenty of help from AI while keeping the information covered. The key is separating the "content" from the "form." Leave the form to AI and fill in the sensitive content only inside the school.
- Pseudonymize: Replace names with symbols, entering it as "Student A did…"
- Generalize: Instead of a specific case, convey only the situation, as in "a diligent student who finds presenting difficult."
- Ask for the form only: With no individual content, ask, "Just build me a template for writing this kind of text."
- Combine afterward: Take the empty template from AI, then fill in the actual student information only in internal school documents.
AI does an excellent job of building the shape of a text even when it knows nothing about the student.
Teach students exactly the same thing: that not typing your own name, a friend's information, or your home address into a chatbot is a basic of online safety. With younger students, the analogy "an adult you don't know might see what you write here" makes it easy to understand.
Key Takeaways
The principle of prompt hygiene is simple: don't type anything you couldn't read out loud in front of a stranger. First, identifying information, sensitive information, other people's information, and confidential material stay out. Second, work around it with pseudonyms, generalization, form-only requests, and combining afterward. Third, teach students the same input safety rules. One line typed in for the sake of convenience can put a person's private life at risk. Before you send the question, the habit of pausing once to ask "is this okay to include?" protects everyone. The busier you are the harder that single breath is, but that brief pause protects both a student's privacy and your own accountability. Calling the list of what must not go in to mind one more time before you type: that is the most basic security rule for a teacher in the age of AI.

Be the first to comment.