Last updated: November 6, 2026

MyTextMate uses one AI model to understand the texts you send. This page explains what it does, what it sees, and how we keep it in check.

What the AI Does

When you text something like "dentist tues 3pm", an AI model reads it and turns it into actions on your own list: add something, show what's coming up, change, finish or delete something, change a setting, or save your feedback. Our code then checks each action and works out the exact date and time before anything is saved.

Every reply says what was done, so if the AI misread you, just text the correction.

What Doesn't Use AI

• Your daily list and the reminders you asked for.
• Your Canvas deadlines, which come straight from Canvas.
• STOP, START, HELP and CANVAS, which are answered without the AI.

What the AI Sees

• The text you just sent.
• Your own upcoming items: what you've texted us, and your Canvas items for the next 30 days.
• Today's date and time where you are.
• Your previous text and our reply, if they were in the last 10 minutes, so a follow-up like "make it 6am" makes sense.

It never sees anyone else's information.

What the AI Can't Do

It has no internet access, tools or plugins. It can't text anyone or see other people's lists. It can only suggest a small, fixed set of actions on your own list, and our code checks every one before it's saved.

The Model

The model is gpt-oss-120b, an open model from OpenAI, run by Cloudflare Workers AI. Cloudflare doesn't keep your texts after answering and doesn't train on them. We never train any model on your information.

Keeping It Accurate

• Any change to the model, or to how it's instructed, must first pass our test suite of more than 100 real-world texts.
• We review recent conversations, with the help of Anthropic's AI assistant, Claude, to find texts the AI got wrong. Each real mistake becomes a permanent test, so it stays fixed.
• Actions happen right away, without a person approving each one, because they only change your own list and every reply shows what was done.
• If the AI ever misbehaves, we can turn it off within minutes. Your daily list, reminders, STOP and HELP keep working without it.

Our practices follow the NIST AI Risk Management Framework. Our privacy notice covers everything else we collect and how long we keep it.

Questions

Email support@mytextmate.com. We answer within 10 business days.