You have some questions on Machine Learning (ML).
SAP Cloud for Customer
it depends. ML heavily depends on quality data.
If your test tenant has a real and recent production data set, then you can do the model quality test in test tenants.
Otherwise you can do only the technical test in the test tenant (whole workflow of ML).
However, you wouldn't be able to assess the model quality based on either poor quality or non-business relevant data created by the testers the test tenant.
Please also refer the doc Machine Learning in SAP Cloud for Customer.
2. Does data leave our single private tenant to get calculated elsewhere?
Everything is happening outside the C4C tenant. The ML solution is not based on ABAP stack.
Only the ML client of C4C is running in c4c abap stack.
For training purpose data is retrieved into SAP ML technology stack (SCP based).
Data is not persisted there permanently. SCP instances are running side-by-side to c4c by respecting the tenant boundaries
(no mixing of different tenants). SCP is part of the C4C data sub-processor list.
3. Does Machine Learning adhere to EUDP?
as explained already, it happens side-by-side SCP ML environment. There is no data sharing between the tenants or external.
EUDP is adhered to.
KBA , LOD-CRM-PM-SLS , C4C Product Management - Sales , LOD-CRM-ML , C4C Machine Learning , How To