Use of AI
Last Updated: September 2026
Photo: Carol M. Highsmith, Library of Congress, public domain
No AI service receives the words you say with Model Communicator.
Model Communicator is made by a very small team, and we used AI tools to help build it. People who rely on AAC, and the families and professionals who support them, deserve to know exactly where AI was involved and where it was not. This page explains it in plain language.
This page grows out of our Frictionless Philosophy: AAC should reach the people who need it with nothing in the way, and that includes nothing between them and their privacy.
The short version: We used AI tools to help write our software and draw our symbols. No AI service receives the words you say with Model Communicator. The web app has no AI that touches your data. The iPad app has one AI feature, smarter word prediction, and it runs entirely on your iPad.
At a glance
How We Used AI to Build Model Communicator
We used two kinds of AI tools to make our apps and this website.
Software
We used Claude Code, an AI coding assistant made by Anthropic, to help write and maintain the code for our apps and this website.
Symbols
The pictures on our communication boards were created with OpenAI's image generation tools, the same technology behind image creation in ChatGPT.
People stay in charge of this work. The vocabulary, the layout of the boards and the way the app supports language learning are decided by our team, led by a practicing speech-language pathologist. We review and test what the AI tools produce, and we replace symbols and fix code that does not meet our standard.
What AI Touches Your Data
Almost none, and nothing that leaves your device. Here is every place, by app.
| Where | AI that sees what you say | Where it runs |
|---|---|---|
| Web app | None | Not applicable |
| iPad app | Smart predictions only (optional) | On your iPad |
| Chrome app | None. It opens the web app. | Not applicable |
No AI model sees your words in the web app. Word prediction uses a fixed table of common word patterns, plus a count of the words you use that is stored only in your browser. It works offline, and your words are not sent to us or to anyone else.
Speech uses the voices built into your browser or device. Those voices come from the maker of your browser or operating system, not from us. If you choose the MoCo Child voice, it runs inside your browser, and the words you speak are not sent anywhere.
Our Concerns About AI
We use these tools with our eyes open. We share many of the concerns that people have raised about AI.
Energy and water
Large AI models take a great deal of electricity and water to train and run. See our Carbon Commitment.
Creative work
Many AI models were trained on the work of artists, writers and programmers who were not asked and were not paid.
Bias
AI systems can repeat and amplify unfair patterns, and they often serve people with disabilities poorly.
Voice and autonomy
In AAC this matters most of all. A communication tool must say what its user means. AI that guesses, rewrites or speaks on someone's behalf can take away the very thing AAC exists to give.
Concentration
A handful of large companies control the most capable AI tools.
Why we used AI anyway
We also want to be honest about why we used AI anyway. Robust AAC apps often cost hundreds of dollars, and that cost keeps communication from people who need it. Without AI tools, a team of our size could not have built a full AAC system and offered it for free. We decided that was worth it. We do not think that decision ends our responsibility, which is why this page exists.
Our Plan: Smaller, Local, Open
Our goal is to reduce our reliance on large, cloud-based AI services. When we are able to purchase our own capable hardware, we plan to move our ongoing development work to lightweight, open-source models that run on that hardware. Smaller models running locally use far less energy, which shrinks our carbon footprint, and they keep our work independent of any one AI company.
We cannot yet afford that hardware, and we will update this page as that changes.
Energy is only part of why we use AI carefully. Our Carbon Commitment explains where our footprint comes from and what we promise to do about it.
Our Commitments
What we promise
- Model Communicator will never speak for you. Prediction only offers words. You choose every word that is spoken.
- We will not send what you say to an AI service.
- We will not use your data to train AI.
- Any AI feature that reads your words will run on your own device and will have an off switch.
- People, including a speech-language pathologist, make the decisions about vocabulary and design.
- We will update this page before anything described here changes.
Questions about how we use AI?
We would like to hear from you, and we will update this page before anything described here changes.
Privacy Policy · Frictionless Philosophy · Carbon Commitment