AI policy

How Alto uses AI

Effective October 10, 2026 · Alto Tech LLC

  • Alto is an AI planning assistant. When you chat with Alto, you are working with AI.
  • It drafts only what you ask for, and nothing reaches students unless you put it there.
  • Alto uses models from Anthropic, with OpenAI as a backup, and an open model run by Baseten for a final review of lesson plans. Neither Alto nor any of them trains AI on your work.

What the AI does

Alto Tech LLC uses AI in Alto to help teachers plan:

  • draft pacing guides, lessons, assessments and student materials for a unit;
  • edit a document when you ask, and check a draft against its standards and your calendar;
  • give each drafted lesson plan a final review that fills gaps, such as a missing expected answer, before you see it;
  • read files you upload, such as a pacing guide or lesson template, to bring them into your unit;
  • suggest standards that fit a unit, and translate student copies of a document;
  • answer in chat, and summarize a long conversation so it can continue;
  • search the web, only when you turn web search on for a chat.

Alto also turns the standards catalog and your own documents into search vectors, so you can find standards and your past work by meaning. Today that model runs inside Alto’s own database. If we switch it to OpenAI’s embedding model, the same no-training terms apply.

You choose when it runs

Using Alto is choosing an AI assistant, and every account knows that from the start. Within Alto, AI works when you ask it to: drafting a day, sending a chat message, importing a file, or preparing materials you selected. It keeps your unit’s materials list in step with documents you change, so it never drifts from your plan. It never publishes, shares or sends anything on its own. Web search and Google Drive stay off unless you turn them on.

Where the AI comes from

Alto doesn’t build its own language models. It uses models from Anthropic, a third party, reached through Vercel’s AI Gateway. When Anthropic can’t answer a request, the same request goes to a backup model from OpenAI. These drafting and editing requests are processed in the United States. OpenAI also runs a quality check on test questions, which sees only the question. Both companies’ terms for this service bar them from training on what Alto sends. Anthropic deletes requests within 30 days, except content its safety systems flag or the law requires it to keep; OpenAI may keep requests for up to 30 days to detect abuse, then deletes them. Alto chooses the model for each task; you never need to pick one.

After a lesson plan is drafted, an open model (GLM 5.3, made by Z.ai) gives it a final review. That model runs on Baseten, also reached through Vercel’s AI Gateway, in the United States. Baseten keeps no copy of these requests and doesn’t train on them, and they never go to Z.ai.

Alto shares your work with Anthropic, OpenAI and Baseten only as needed to carry out your request, and with no one else for AI purposes. Every provider that handles your information is listed in our Privacy Policy.

No training on your work

  • Alto does not train, fine-tune or otherwise improve any AI model with your work, your prompts or your students’ information.
  • Anthropic, OpenAI and Baseten don’t train models on Alto’s requests.
  • Information from Google is never used to create, train or improve any AI model.
  • Each model was trained by the company that made it, on its own data, before Alto used it. Nothing from Alto is part of that.

Alto gets better through how it is built: the instructions, the official standards catalog, and the checks it runs on every draft. We measure quality with test units we write ourselves, not with your work.

What the AI sees

Each request carries only what that task needs: what you asked, and the context that makes the draft fit your class. That means the unit’s standards and dates, class needs you described, documents in the unit, your teaching preferences and lesson template, and files you uploaded to the unit. It never includes student names, because Alto never collects them. Anything you upload that other people wrote, or that Alto found on the web, is marked so the AI treats it as material to read, not as instructions to follow.

You review, you decide

  • AI drafts can be wrong. Alto checks every draft against its standards, dates and format and shows you what needs attention, but read it before you use it.
  • Standards come from the official Georgia catalog, and every draft names its exact codes so you can check them.
  • When Alto proposes a change to a document you already have, it shows the change for your review instead of rewriting your work.
  • IEP goals and supports are drafts for the IEP team. Alto never makes educational decisions about a student.

People and safety

Nobody at Alto reads your conversations or documents unless you ask us for help, to investigate abuse or a security problem, or when the law requires it. We watch usage and costs, not content, to keep Alto running. Alto is built to produce material suitable for school; tell us at once if it doesn’t, and we’ll fix it.