Policy

Data Safety Guide

Practical data safety rules for AI-assisted workflows.

Last updated: May 13, 2026

Do not paste unnecessary sensitive data

Before using an AI assistant, remove information that is not required for the task. Replace real names, emails, account numbers and private identifiers with placeholders when possible.

Classify the task

Separate low-risk tasks such as grammar improvement from high-risk tasks such as legal interpretation, medical information, security configuration or financial decisions. High-risk tasks require expert review.

Keep a review log

For business workflows, save the prompt version, source material, output, reviewer and approval date. This helps teams understand what was generated and what was changed by humans.

Check retention settings

Different platforms may have different settings for chat history, data retention and enterprise controls. Users should review official settings before entering confidential material.

Use secure channels

Do not send private files or credentials through unsecured forms. Production systems should use HTTPS, access controls, least privilege and secure storage.