Ops & admin automation
Hand repeat admin (receipts, logs, notices, handoffs) to workflows your agent can trigger.
It depends on the plan, settings and contract, not the brand. Business and API plans often offer stronger data terms than free consumer apps, but you must check each vendor's current retention and training terms, your client contracts and privacy law. Until reviewed, keep client data out.
Approval depends on the exact product, account, settings, contract and data. A paid personal subscription is not automatically a business data agreement.
A documented decision gives staff a practical rule and makes exceptions visible. It can reduce accidental disclosure; it cannot guarantee legal compliance or prevent every breach.
“No training” is not “no storage.” Turning off training does not delete existing chats. OpenAI and Anthropic both describe feedback exceptions, so include feedback settings in the review.
Plugins, MCP servers and other connected services may receive selected input and apply separate terms. Check what actually leaves the workspace and who can retrieve the resulting records.
Special categories of data and regulated work may require additional safeguards or agreements. This is an orientation checklist, not legal advice or a declaration that a particular client dataset is safe to upload.
Snapshot from official sources on Oct 18, 2026.
| OpenAI business | No training on Business, Enterprise, Edu or API data by default; retention controls depend on eligibility. No-training does not mean no retention. [3] |
| OpenAI personal | Turning off model improvement affects new conversations; feedback can be an exception. Codex environment context has a separate setting. [4] |
| Anthropic commercial | No training by default; feedback or explicit opt-in can allow it. Team/Enterprise owners can disable chat-rating feedback. [1] |
| Anthropic consumer | Free, Pro and Max follow consumer rules, including their Claude Code use; improvement opt-in, safety-review and feedback exceptions apply. [2] |
| Qualifying Google Workspace editions | Content is not human reviewed or used for model training outside the domain without permission. Retention differs by product and admin settings. [5] |
Hand repeat admin (receipts, logs, notices, handoffs) to workflows your agent can trigger.
Not tool-specific. See Agent-ready for connector support by tool.
Levels come from each tool's recorded, sourced compatibility. How the levels work.
Practitioner views are restated in our words and labeled as views, not measured results. Product facts come from official sources dated above.
IonBin's wiki isn't user-edited. Corrections go to our Directory Curator, who checks them against primary sources before anything changes. Spot something wrong or out of date on “Is it safe to put client data into AI tools?”? Email us with a link to a primary source.
Suggest an edit by email or write to contact@ionbin.com