AI in Origin
What AI does in Origin
Origin uses AI for a few specific jobs:
- Drafting courses, lessons and quizzes from the PDF or Word files you upload.
- Writing short summaries, quick-read and easy-read versions of your lessons.
- Writing Munin’s short review questions from your lessons.
- Scoring some practice answers, so Munin knows when to schedule the next review.
- Reading lessons aloud as a voice-over, and voice practice where learners explain a topic out loud.
- The Origin AI assistant for admins, which answers questions and proposes changes you confirm.
The models we use
Origin uses AI models from three providers, through their business APIs:
- Anthropic (Claude) for all written work: drafting, summaries, review questions, scoring written answers and the admin assistant.
- OpenAI for voice practice only: turning speech into text, spoken replies and live voice conversations.
- Google Cloud Text-to-Speech for lesson voice-overs.
Providers release new model versions often, and we move to them. We update this page when we add or change a provider.
What a person checks
- Courses and lessons you build with Origin AI from your material wait for your approval before learners see them.
- Summaries, quick-read and easy-read versions stay hidden from learners until a teacher has reviewed them.
- Munin’s review questions are written by AI from your approved lessons and checked automatically against the answer key, not by a person.
- Question banks you generate in bulk are added to the module when you start the batch. Check them in the quiz editor.
- The admin assistant never changes anything on its own. It proposes a change, you confirm it, and it can be undone.
AI and learners’ scores
Some practice formats are scored by AI: written explanations, short recall answers, scenarios and voice practice. The score only decides when Munin schedules the next review of that skill.
AI scores do not affect grades, certificates or pass and fail. Origin does not use AI to decide admission or course level, and does not use AI to monitor learners during tests.
What data the models see
- Course material you upload goes to Anthropic, so it can be turned into lessons.
- Learners’ written practice answers go to Anthropic to be scored, without their names.
- Learners’ voice in voice practice goes to OpenAI, only if your organisation has switched voice practice on.
- The admin assistant can look up people’s names and emails to answer an admin’s question, and those go to Anthropic.
- Lesson text goes to Google for voice-overs. Names are removed from the summaries managers receive.
Anthropic and OpenAI state that they do not use API data to train their models by default (Anthropic, OpenAI). Both may keep API data for up to 30 days for abuse monitoring.
Anthropic processes requests on global or US infrastructure; it does not offer EU-only processing today. Voice practice with OpenAI is also processed outside the EU. Origin keeps a log of AI prompts and replies for up to 180 days, for quality and safety.
Your controls
- AI drafting of course content can be switched off for your organisation.
- Voice practice stays off until you switch it on.
- Munin’s reviews and Hugin’s practice stay off until you switch them on.
- AI marking of hand-ins is not switched on for any customer.
Ask us to switch any of them on or off, or to go through the settings with you.
The EU AI Act
Drafting teaching material is not one of the education uses the AI Act lists as high-risk. Using AI to evaluate learning outcomes, including to steer the learning process, can be. After the 2026 amendment, those high-risk rules apply from 2 December 2027 (Annex III, European Commission).
Origin’s AI scoring of practice answers influences when a review comes, so we are assessing it against those rules before that date. AI marking of hand-ins stays switched off until that assessment is done.
Since 2 August 2026, the Act also requires that people know when they are talking to an AI and that AI-generated audio is marked. We are aligning Origin’s labelling of AI voices and AI practice partners with those rules.
This is our reading of the Act, not legal advice. Your own obligations as a school or training provider may differ.
Questions
If your IT team, data protection officer or procurement has questions about AI in Origin, call +46 76 426 26 60 (Monday to Friday, 09:00–18:00 CET) or write to us through the help center. We send the data processing agreement and the list of sub-processors on request.