Before You Prompt: Why Privacy Belongs in AI Literacy

It is 11:47 p.m., an essay is due in the morning, and a student opens an AI chatbot for feedback. The draft includes her full name, her school, a story about a family argument, and enough detail to identify everyone involved. She pastes the whole thing into the chat box. It feels private.

That feeling is powerful. A chatbot answers one person at a time and remembers the thread, while its replies follow the easy rhythm of a conversation. The screen looks closer to a private notebook than a public website. Yet the student's words are entering a service governed by a privacy policy and terms of use, documents that can explain how prompts are collected, retained, reviewed, shared, or used to improve a system. Most students meet those documents through one checkbox, usually while trying to get past it.

Here, Termcon and SAIL meet. SAIL, short for Students for AI Literacy, runs student-led workshops that explain the mechanics and limits of AI. Students also examine bias and practice using the technology responsibly. The organization has already taught more than 600 students across five chapters. Termcon focuses on the agreements people accept online and turns dense terms into language a normal person can examine. That overlap is easiest to see when a student is deciding what to type into an AI chat box.

Students are making that decision constantly. A 2026 Pew Research Center survey found that 64% of American teens use AI chatbots, 54% have used them for schoolwork, and 12% have asked them for emotional support or advice. Each use can carry a different kind of information. A math question might reveal almost nothing about its author. When the topic is a college essay, recommendation letter, health question, or late-night conversation, the prompt can reveal names, routines, fears, grades, family relationships, and plans.

UNESCO has made this exact problem part of AI education. Its 2024 AI Competency Framework for Students includes privacy within the ethics of AI, one of four dimensions in a framework containing 12 competencies. One exercise asks students to imagine entering personal information into a prompt so an AI system can draft a recommendation letter, then question whether that information could be disclosed. Drafting help sounds ordinary, so students may overlook that a recommendation-letter prompt can carry a bundle of personal data.

The uncertainty reaches beyond homework. In September 2025, the Federal Trade Commission opened an inquiry into seven companies that offer consumer-facing AI chatbots. The agency asked about the use or sharing of personal information from conversations. Its order also covered age restrictions, disclosures about data practices, and safeguards for possible harms to children and teens. The inquiry did not declare that every chatbot follows the same practice. Since policies and settings differ across services, the FTC's questions point students back to the document they skipped when creating the account.

So what should a student look for? Start with what happens to the words they enter. Does the service use prompts to train or improve its models, and how long can it keep chats? Check for human review. Then find out whether the company shares conversations with contractors or other companies and what controls exist for deleting history or opting out. In a workshop, students could use Termcon to pull out the relevant clauses, then use SAIL's discussion format to decide whether the tool fits a school task.

Data minimization comes next. The formal phrase describes a pretty simple habit: give the tool only what it needs. A student asking for comments on a resume can replace a phone number and home address with brackets. Someone revising a personal essay can change a sibling's name, remove the school, and describe a medical issue without attaching a diagnosis to an identifiable person. For a recommendation-letter draft, "a student in my AP Biology class" may work as well as a full legal name. Small edits can preserve the usefulness of the prompt while lowering the cost of a mistake.

Schools can teach this in one class period. Give students a fictional prompt containing a name, school, birthday, and private family detail. Ask them to mark what the AI needs, redact what it does not, and inspect the tool's policy for retention, model training, human review, and deletion. Then let groups compare what they found. SAIL's near-peer format makes that conversation less stiff because students will admit how they use AI to another student in ways they may avoid saying to a teacher. Termcon can help them work through the policy language without pretending that a forty-page agreement is reasonable homework.

Plagiarism leads most school AI rules. Those rules matter, but academic honesty covers only what a student produces, while privacy asks what the student gave away during the process. A perfectly original essay can still expose a friend's secret. The same risk appears when a factually correct study guide begins with a prompt containing a student's disability status or class schedule. Schools need room for both questions as they write AI rules for the coming year.

At 11:47 p.m., legal text is an easy thing to skip. Take one minute. Remove identifying details and check the privacy controls before posting. If the remaining prompt would feel unsafe in the hands of a stranger, close the tab. The student can still get feedback on her essay. Her name, school, and family argument can stay in the document on her own laptop.