When sensitive data cannot be sent to a third-party API, data sovereignty becomes a deployment decision: keep inference within the organization’s infrastructure or an approved jurisdiction. Fine-tuning alone does not provide this protection; where the model runs determines whether data stays inside the trusted boundary. section 1.3
Common deployment choices include:
- Run a model on-premises or in a private cloud. section 1.3
- Use a managed service in the customer’s cloud region with data-residency and no-training commitments. section 1.3
A regulated deployment
A team can self-host Qwen3-4B with retrieval and leave its weights unchanged to meet a regulator’s boundary requirement. If the small model needs improvement to do the job, adaptation may help; it is not what keeps the data in-bounds. An in-house deployment may cost more at low volumes, yet remain usable when a frontier API is not. section 1.3 section 1.3


