- Reduce personal data before using AI.
- Distinguish local AI, cloud providers, and data processors.
- Create a minimal checklist before automating client tasks.
Minimal checklist
- Purpose: what the AI is used for.
- Data: which fields go in and which are removed.
- Provider: local, self-hosted, or external service.
- Access: who can see input, output, and logs.
- Retention: how long you keep results.
- Human review: which actions require approval.
Before sending data to a model: 1. Remove unnecessary fields. 2. Replace names with internal IDs when possible. 3. Do not include sensitive data unless clearly necessary. 4. Do not allow automatic sending without review. 5. Record what the system did and who approved it.