The best way I can put it is: You can use and train AI on your stuff, but you can’t on our stuff.
So, when you wrote this
Allowed:
- local automation
- own-project research
- synthetic data from own scenes
- RL/simulation/testing inside own projects
Not allowed:
- scraping Unity services
- cloning Unity documentation or Asset Store data
- training models on Unity proprietary materials
- bypassing subscriptions, rate limits, or security
- using Unity data for competitive replication
you got it, though I will add “training on the Unity Runtime” to your list of Not Allowed