How CuraCourse uses AI.

A note from our founders

We’re educators, and we build for teachers and students. AI showed up in classrooms faster than anyone could write a policy for it. Some instructors want it everywhere. Others want it nowhere near their course. Both are reasonable, and CuraCourse works either way.

Our AI works from the content you write and follows the settings you choose. It’s built to help students think, not to think for them. Student work is never used to train AI models. We’re studying what helps students learn with University of Rhode Island researchers, and we change the product based on what we find.

Joe Mazzone and Brian Chung, Co-Founders

Our principles

Grounded

The AI works from your course content, so it uses your material, notation, and order.

Instructor controlled

You decide where AI shows up and how much it can do. The tutor is off until you turn it on for a lesson.

Learning first

The tutor helps with questions and hints. It’s built not to give out answer keys, hidden tests, or full solutions to graded work.

Private

Student work is never used to train AI models, ours or anyone else’s. Model requests only go to zero data retention providers.

Visible

Tutor chats count as coursework, so you and your course staff can read them.

Evidence based

We’re studying how AI helps students learn in real courses with University of Rhode Island researchers. What we learn shapes the product.

You set the limits

Each lesson has its own tutor setting, and students only get the help you allow. The tutor isn’t available on exams. It comes with the Core and Pro student plans.

Where AI appears

Each feature has its own switch. Turning on the tutor doesn’t turn on AI Chat or AI grading.

CuraAI for course authors

How tracked changes work

Write and revise lessons, build custom activities, and work with your course material.

Controls. You start the conversation and decide what to publish. Its edits show up as tracked changes you can accept or reject.

What the AI sees. Your prompts and chat history, the lesson or book content it’s working on, and any files you attach.

The lesson teaching assistant

Set up the tutor

Helps students work through a lesson and talk about their own work.

Controls. Off by default. You turn it on lesson by lesson and pick a level from 1 to 3. At Level 3, students accept or reject each suggested code edit.

What the AI sees. The lesson and the chat so far. At Levels 2 and 3, it also sees the student’s work and any feedback you’ve released.

AI Chat activities

Set up an AI Chat activity

Students chat with a character or task you set up.

Controls. You add the block and set its mode, instructions, message limits, attachments, and content filters. Fully simulated mode uses replies you write. AI and hybrid modes can use a model.

What the AI sees. Your instructions, the conversation, and any attachments you turned on, when a model reply is needed.

AI grading and feedback

Set up AI grading

Scores work and writes feedback using AI rubric criteria, including on practice checks.

Controls. You set the criteria, points, instructions, and what each score means. It can grade automatically. Try it on a few answers you’ve already graded before you use it with a class.

What the AI sees. The question, the student’s answer or files, and your rubric criteria.

Use as much or as little AI as you want

Lessons, choice questions, blanks, ordering, graphs, code tests, and manual grading all work without AI. Leave the tutor off when it doesn’t fit the assignment.

AI can be wrong, incomplete, or biased, so try what it makes before your students do. Tutor levels and content filters steer it, but they can’t guarantee every reply is appropriate or never gives away an answer.

What happens to your data

AI requests run on outside model providers. Which one depends on the feature, the model, and availability. Every route, including backups, requires ZDR. If no ZDR provider is available, the request fails rather than going somewhere else.

Student work is never used to train AI models. For schools, we act as a “school official” under FERPA, and we follow COPPA for students under 13. We sign data privacy agreements on request, and schools can ask us to delete their data.

Only share what the task needs. Don’t enter passwords, unrelated personal details, or material your school hasn’t approved for use with outside AI services.

Research

We’re working with researchers at the University of Rhode Island on an IRB-approved study of how AI can help students learn in real courses. What we learn shapes the product.

Questions

For students and families

Your instructor decides whether your course uses AI. Follow your course’s AI rules, and ask your instructor if you’re not sure.

Your answers and tutor chats are coursework, and your instructor and course staff can read them. AI feedback can be wrong. If you disagree with an answer or score, talk to your instructor.

Reviewing CuraCourse for your school?

Tell us which features your instructors plan to use and what your review needs, such as consent, data retention, or procurement. We’ll walk you through how the data moves before you roll it out.