Does AI Learn From My Child's Data? In-Context Learning vs. Model Training, Explained for Parents
When parents hear that a tutoring service uses AI, the same worry comes up: "Is the AI learning from my child? Will my kid's work end up inside some giant model?"
It's the right question to ask. The answer depends on which of two very different ways the AI uses information. Once you know the difference, you can judge any AI product for yourself - including ours.
The short version
- Model training permanently changes the AI. Data used in training can shape what the model says to everyone, later.
- In-context learning means the AI reads information for one task - like you reading a note - then answers. The AI itself doesn't change, and it doesn't remember the note afterward.
- pAIrmatic uses AI only in the second way, with model training turned off at our AI providers. Your child's information is not used to train their models.
Way 1: Model training - how an AI is built
An AI model starts as a huge set of numbers, often called its weights. Training is the process of adjusting those numbers by showing the model enormous amounts of text, over weeks or months, on large clusters of computers.
Think of it like years of school. Everything a student studies becomes part of what they know, and they carry it into every future conversation.
That's why training data matters for privacy. If someone's personal information were used in training, it would become part of the model's permanent "knowledge," shared by everyone who uses it. You can't easily pull one person's information back out.
So the key question for any AI product is: "Is my data used to train the model?" Some consumer chat apps may use conversations to improve their models, depending on the app and your settings. For pAIrmatic, the answer is no: we use our AI providers with model training turned off.
Way 2: In-context learning - how pAIrmatic uses AI
The model's "context" is the information included with a single request: the instructions, plus any material the AI needs for that one task. The model reads it, uses it to produce an answer, and that's it.
Think of an open-book exam, or a substitute teacher handed a lesson-plan folder. The substitute reads the folder, teaches the class well, and hands the folder back. Reading it didn't change who they are, and they don't take the folder home.
AI models work the same way. The technical term is stateless: each request starts from a blank slate. The model has no memory of your child between requests. If we want the AI to know something about your child's progress, we have to include it again, every single time.
A real example: planning your child's next session
Here's what one in-context request looks like at pAIrmatic, using a made-up student. Before a session, we send the AI something like this:
Instructions: Plan a 60-minute math session. Start with the lowest skill that's still shaky.
Student: Grade 7.
Check-in results: Solid on one-step equations. Shaky on negative numbers - adds them as if two negatives make a positive.
Last session: Practiced adding integers on a number line.
Homework: 6 of 8 correct. Both misses were subtracting a negative.
A plan for the tutor: start with five warm-up problems on subtracting negatives, rebuild the rule on a number line, then move to two-step equations that include negatives.
Then the request is over. The model doesn't keep the profile, doesn't learn from it, and doesn't get better at anything for anyone else. Next week, we send an updated profile, and the model reads it fresh.
Notice two more things:
- The AI plans. A person teaches. The plan goes to your child's tutor, a real, interviewed person who runs the live session and adjusts to how your child is actually doing that day.
- The "memory" is ours, not the AI's. The student profile lives in our own private database, not inside the model. That means we control it, follow our deletion schedule, and delete it if you ask.
Side by side
| Model training | In-context learning | |
|---|---|---|
| Analogy | Years of school | Reading a note for one task |
| Changes the AI? | Yes, permanently | No |
| Remembered later? | Yes, in the model | No - each request starts blank |
| Who controls the data? | Baked into the model | Stays in the company's own records |
| pAIrmatic uses it? | No - training turned off | Yes |
What pAIrmatic uses AI for
As described in our Privacy Policy, we use AI tools to:
- Write session summaries and plan the next session.
- Review homework your child sends in.
- Help score answers on the check-in.
- Check session transcripts for shared contact details and safety concerns. Any match goes to our founder, who reviews it.
AI never teaches your child. A tutor runs every live session.
How your child's information is protected
- Training is off. Our AI providers don't use your child's information to train their models.
- Only what's needed. We send the AI only what a task requires. On the check-in, we ask students for initials, never full names.
- Short retention. Session recordings, transcripts and homework photos are deleted within 15 days, unless a safety concern requires keeping one.
- Your control. You can ask us anytime to review, correct, or delete your family's information at info@pairmatic.com.
- Consent first. Sessions are recorded only with your written consent.
Read the full details in our Privacy Policy and our Parent's Guide to Session Safety.
Questions to ask any AI tutoring company
Use these with any service, not just ours:
- Is my child's data used to train AI models - yours or your providers'?
- Does AI teach my child, or does a person?
- What exactly do you send to the AI?
- How long do you keep recordings and transcripts?
- Can I see and delete my child's information?
A trustworthy company answers all five plainly. If you'd like to ask us directly, email info@pairmatic.com.