How we use AI

What the AI does

• Reads the handouts you upload and lists the skills they teach.
• Writes the practice activities and questions for those skills.
• Sorts some wrong answers into kinds of mistake (for example "wrong ending" or "wrong helper verb").

Every activity built by AI carries a small "Written by AI" label.

What the AI doesn't do

• It doesn't grade you. Our own code checks every answer, the same way every time.
• It doesn't decide what you study next. Our code does that, from your answers.
• It never writes your homework or your French for you.

How questions are checked

• Our code checks each activity's format and that its answers can actually be reached.
• A second AI checks each answer key before the question reaches students.
• Teachers who build units for a class review them before publishing.
• You can report any question with "Report this item". Two student reports, or one from the teacher, hide the question for the class until someone checks it.

Even so, mistakes can get through. Cahier is a study tool, not your teacher's answer key.

Which AI, and what it sees

Cahier uses Claude, made by Anthropic. When a unit is built, Claude sees the handouts and the note for that unit. When it sorts mistakes, it sees the question and your wrong answer (up to 80 characters), with no name, email or account id. Anthropic's commercial terms say it does not train its models on this data. It keeps it only for a short time, except that content its safety systems flag may be kept for up to 2 years.

While Cahier is being built, its developer also tests Google's Gemini with their own test handouts. Gemini is never used for students' accounts.

Nobody, including us, trains AI on your data. More in the privacy policy.

Questions

Write to support@studycahier.com.