Does an AI tool learn from what your students type?

It depends on the company and on the plan. Some companies use what people type and upload to train future versions of their AI. Some say clearly that they do not. Others do it unless someone turns a setting off. The company’s own pages for the plan your school uses will tell you which, and the default setting, the one a new account starts with, is the one that counts.

What training means

When a student types into an AI chat, the AI does not change in that moment. Training happens later, when a company takes saved conversations and uploads and uses them to build or improve a model. If a student’s essay or question goes into that process, it helps shape a product the company keeps using long after the lesson ends.

That is why this question matters even when nothing seems to go wrong. A student may type something personal without thinking, such as a worry, a family detail or their own name. Deleting the conversation later may remove the saved copy, although some companies keep deleted data for a time or in backups. It may not undo training that has already happened.

Training is not the same as storing or remembering

Three different things can happen to what a student types, and it helps to keep them apart.

The company stores it. A company may keep what you type for some time, even if it never trains on it. Where it is kept, and whether it can be deleted, are separate questions, covered in our article on where student data goes.

The tool remembers it. Some AI tools have a memory feature that keeps notes about a user from one chat to the next. That changes what the tool says to that user, but it is not the same as training the company’s model.

The company trains on it. This is the question in this article: does what people type become part of how the company builds its AI?

A company can store data without training on it. So a tool that does not train is still worth checking on the other questions.

Why the default setting matters

Some tools that train on what people type also offer a setting to turn it off. That sounds reassuring, but a setting only protects the people who change it. A student with a new account, or a teacher trying a tool for the first time, starts with whatever the company chose.

So our check looks at the default. If training is on until someone turns it off, the tool can still be useful, but the school needs to know, and ideally turn it off for every account before students use it. Some education plans let a school administrator do that for everyone at once.

How to find the answer

Start with the plan. The free plan and the education plan of the same product can have different rules about training, so make sure you are reading the pages for the one your school uses.

Then read three places. In the privacy policy, look for wording about using content to improve the service or to develop models. In the help centre, look for a data or privacy setting, and check whether it starts on or off. In any education terms or data agreement, look for a line about whether student content is used for training.

Some tools add an AI feature to software your school already uses, such as a writing or quiz tool. Behind the scenes, that feature may send what people type to another AI company. The tool’s pages should say whether that other company can train on it.

If none of these pages answers the question, that is an answer too: it is not confirmed. A short email to the company asking only that question is a fair next step.

What the answer means for the check result

Our tool pages record one of four answers for this fact.

“No” means the company says it does not train on what people type in that plan. This meets our green rule.

“Yes, unless you turn it off” means training is on by default. The best result the tool can get is “Check before student use”, even if your school turns training off, because we check what happens when nobody changes anything.

“Yes, and you cannot turn it off” gives the result “Not for student data”. Staff can still use such a tool for planning with no student information in it.

“Not confirmed” means the company’s pages do not say. Unless another rule already makes the tool “Not for student data”, the result is “Not confirmed yet”.

What teachers and parents can do now

Whatever a tool’s answer, the simplest protection is to keep personal details out of it. In Victorian government schools, the department’s generative AI policy (read 25 September 2026) requires schools to direct staff and students not to put personal information about students or staff into generative AI tools. It also asks schools to prefer tools that do not use the data to train an AI model. Other schools can check their sector’s or state’s rules.

For teachers, that often means using a placeholder instead of a name, as our article on keeping student names out of report comments shows. It also means checking the training setting on your own account, and telling your school if you find it switched on.

For parents and carers, it is fair to ask the school which AI tools your child uses and whether those tools train on what students type. Our article on questions to ask about AI at your child’s school suggests a few more.

As we publish tool checks, each tool page in our tool directory shows the training answer and its source.