Data processing agreement
1. Roles
For the data you entrust through projects, repository content, artifacts and product data traversing runs, Milin acts as your processor and follows your documented instructions. For your account data, identity, billing and portal usage, Milin is the controller.
2. Scope of processing
Project data is processed solely to execute the runs you assign: reading the inputs you connect, producing deliverables, and verifying them. It is not used for any other purpose, and never to train models.
3. Sub-processors
Hosting is provided by Google Cloud within the European Union (europe-west). Your chosen AI model provider (Anthropic Console, AWS Bedrock or Google Vertex) is engaged by you, through your own keys. The current sub-processor list is published and kept up to date at general availability.
4. Security
API keys are stored encrypted in a per-workspace vault, never logged, and never shown again after entry. Runs execute with least privilege per brick, in ephemeral containers, and no key is ever reused on your behalf.
5. International transfers
Platform data stays in europe-west. When a run calls your model provider, your code and inputs are transmitted to that provider under your own contract with them, which may process outside the EU. This is stated clearly at onboarding.
6. Retention and deletion
You can export everything or delete your mill on request. Run logs, including itemized inference costs, are retained for audit and made available to you. The definitive DPA is finalized with legal counsel before the first paid contract.