Work AI data hygiene checklist
Use with privacy and data hygiene at work. This records a scope check, not legal compliance or new permission.
Before any submission, upload, indexing or connection
- Exact tool, account/tenant, feature and task: ______
- Relevant policy or authorized decision and owner: ______
- Input data category and minimum needed fields: ______
- Vendors, connected recipients and output destination: ______
- Training, feedback, access, retention and deletion conditions: ______
If permission or a necessary control is unresolved, stop the affected use and ask the responsible owner. A paid plan, training-off setting, temporary mode or local interface does not establish approval.
Prepare the input
Inspect the whole attachment, including comments, hidden sheets and metadata relevant to disclosure. Use synthetic data where structure is enough. Remove credentials and unnecessary records. Changing names or rounding figures does not establish anonymity and can invalidate the task. Keep originals under their existing records policy.
Check the output and its destination
Verify facts, scope and commitments. Use required review and approved recipients. Any permitted automation needs explicit scope, tested controls, an accountable owner and a stop route.
If a boundary may have been crossed
Stop the affected use. Record tool/account/time/data category through the approved incident channel without copying the sensitive material again. Notify the designated responder promptly, preserve evidence and use the authorized credential revocation/rotation route if relevant. Follow responsible instructions on notifications and deletion.
Recheck when the account, feature, model provider, input category, connection, destination or purpose changes.