AI privacy settings checklist: turn off what you do not want trained on

What it is good at

• Covers the three tools most designers actually use day to day, not a generic list.
• Separates the training toggle from the history toggle, which most guides blur together.
• Gives you a repeatable five-minute audit you can run once a term.
• Flags the settings that a work or school account takes out of your hands entirely.

Where it falls short

• Menu names and locations move as tools update; check the exact wording in your account before relying on this list.
• Turning off training does not retroactively remove content already used, and does not delete your chat history by itself.
• Business, education and enterprise plans often have different defaults set by an administrator, not by you.

How to use it for creative work

1. 01

   Audit ChatGPT first

   Click your profile icon, choose Settings, open Data Controls, and turn off Improve the model for everyone. Signed-in web, mobile and voice mode all share this one switch, so you only need to do it once per account.

2. 02

   Audit Gemini

   In the Gemini web app, open Settings and help, then Activity (or go to myactivity.google.com/product/gemini). Turn off Keep Activity if you do not want conversations, uploaded files or Live audio and video saved to your Google Account.

3. 03

   Audit Claude

   Open Settings, then Privacy, and look for the toggle that governs whether your conversations can be used to improve the model. Free and paid consumer plans differ here, so check the exact wording rather than assuming it matches another tool.

4. 04

   Check what a work or school account controls

   If you use a Google Workspace, ChatGPT Business/Enterprise/Edu or Claude for Work login, your administrator sets defaults for you, and you may not see the toggle at all. Ask whoever manages the account what the retention policy actually is before assuming you are opted out.

5. 05

   Delete the backlog, then repeat quarterly

   Toggling training off only protects future conversations. Go into each tool's activity or chat history and delete anything containing client names, unreleased designs or personal data, then put a recurring reminder in your calendar to redo this check every three months.

Ready-to-use prompts

Before I paste anything into you, tell me: is this conversation used to train your model by default on my current plan, and where in settings would I turn that off if I wanted to?

I want to ask you about [project], but I need to remove anything identifying first. Read this brief back to me and flag any client name, brand name, unreleased date or personal detail I should redact before I continue: [paste brief].

Write a one-page AI data policy for a [small studio/freelance team] of [number] people, covering: what may be pasted into public AI tools, what must never be pasted, which tools are approved, and who checks settings each quarter.

I am about to sign up for [tool name] for my business. List the ten questions I should ask their privacy or support page about data retention, training use and deletion rights before I upload client work.

I am reviewing my AI tool settings today. Walk me through, tool by tool for [ChatGPT/Gemini/Claude, list the ones you use], exactly what a default account keeps, for how long, and what changes when I turn training off.

Questions people ask

Does turning off training delete my past conversations?

No. It stops new conversations being used to improve the model, but existing chat history stays in your account until you delete it separately [1](https://help.openai.com/en/articles/7730893-data-controls-faq).

Is a business or education AI account automatically private?

Often the default is not to train on your inputs, but retention and access rules are still set by the administrator, not by you, so ask rather than assume [3](https://help.openai.com/en/articles/8983130-how-does-chatgpt-use-my-data).

Should I ask a lawyer about this?

For anything involving client contracts, NDAs or regulated personal data, a data protection or IP solicitor should confirm what your specific agreements allow before you paste that material into any AI tool.