Responsible AI
- Human plan review before generation, on by default
- Uploaded material treated as untrusted, with defences against hidden instructions and guarded fetching of web pages
- Automated checks on answer keys and question quality
- The model used is recorded for every AI-generated course