An enterprise customer of your code assistant requires that their source code is never mixed with other customers' data and is never used to train models. You are building an embedding index over customer repositories for context retrieval. Which design satisfies this?

1. A Use tenant-scoped indexes with per-tenant keys, enforce tenant ID at query time, and default to no training on customer code.
2. B Use one shared global index for all customers, since embeddings are irreversible vectors and cannot leak the original source code.
3. C Keep a shared index but hash file paths and identifiers before embedding so other tenants cannot recognize matching code chunks.
4. D Store code only in the customer's region and train a shared model on it, since regional residency satisfies the isolation requirement.