Here are a few recommendations for how a small business client could prepare and process data before granting AI access:
- The client should store the raw data files in a secure cloud storage system that they fully control access to. Popular options include Google Drive, Dropbox, Box, etc.
- If there is sensitive information in the data that is not required for the project, the client should remove or redact that information beforehand. This includes personal identities, contact information, etc.
- The client may want to create copies of the data files to provide access to, keeping the original datasets untouched.
- It is advisable to grant the AI temporary read-only access to the cloud folders for the duration of the project. Rescind access once the project is complete.
- The AI operator would not need direct access to the raw datasets. The client grants system access only, without human oversight.
- For access, encrypted keys can be used, or time-limited presigned URLs if using a system like AWS S3.
- The client should verify outputs do not expose any private source data before usage.
The client retains full control and ownership of data throughout the process. Let me know if you need any further guidance on best practices for data preparation and access control!