Processing of uploaded materials

When you upload a file to build your own course from it, Threshold (“we”) passes its content to the services below. Only to build the course: recognise the text, draft a plan and generate lessons, cards and checks. Your name, email and other account data are not sent to these services.

1. Who receives the material

• Text-recognition providers — Mistral AI (France) and, when it is unavailable, LlamaIndex (the LlamaParse service, US). They extract text, tables and structure from PDF and DOCX files. They do not train models on your file.
• AI model providers — Anthropic PBC (US) and OpenAI (US). They read the material in parts: find the topic, draft the course plan, write lessons and check them. They do not train models on your file.

We will send you the providers’ addresses on request at support@threshold.day.

These services’ servers are in the EU and the US, that is, outside the country where some of our users live. If the list changes, we will update this page before a new service receives any material.

2. What not to upload

Do not upload documents containing other people’s personal data: names, contacts, health information. Upload only material you have the right to use.

3. How long it is kept

The file and its fragments are kept while the course exists. When you delete a course, they are erased within a day. A plan you have not approved is deleted after 7 days together with the file.

4. Your consent

The checkbox at upload is your consent to these terms. We store it with the date and the version of this text. Questions: support@threshold.day; how we handle account data is described in the Privacy Policy.