top of page
Search

How to change the privacy settings, so you’re not training the model.

  • jenkerrmarsch
  • Nov 7
  • 3 min read
ree

This is one of the biggest fears of businesses with proprietary information. Because as you use these AI models, you’re training them on your systems, etc. ideas concepts. Which then means that you are coming behind you pretty much learn that much faster if they’re using the same model. You can stop this by changing the settings, creating privacy settings in your account.

Here are the step-by-step instructions on how to do this and this was done with ChatGPT!


TL;DR (What changes right now)

  • Personal ChatGPT accounts: Turn off “Improve the model for everyone” in Settings → Data Controls. Your chats won’t be used to train future models.

  • ChatGPT Business/Enterprise/Edu: By default, OpenAI does not use your org’s data (inputs/outputs) for training. Confirm this in your admin console and share the policy with your team.

  • OpenAI API: Data sent via the API is not used for training by default (unless you explicitly opt in).


Step-by-Step: Stop Training on Your Personal ChatGPT Account

Use this if you’re on Free/Plus and using your personal login for work.

  1. Open ChatGPT (web) and sign in.

  2. Click your profile icon → Settings.

  3. Go to Data Controls.

  4. Toggle “Improve the model for everyone” OFF.

    • Result: Your conversations are not used to train ChatGPT going forward. (Your chats still appear in your history unless you also turn off chat history.)

Optional but smart

  • Turn off Chat History (web/mobile) if you don’t want your new chats stored in your account. On iOS: ChatGPT app → Settings → Data Controls → Chat History.

  • Clean up sensitive threads you already created: Delete them from your sidebar (note: deletions can be subject to a retention window).

  • Export a copy of your data before you tidy up: Settings → Data Controls → Export.


Step-by-Step: If You’re on ChatGPT Business (recommended for teams)

Best choice for 1–10 person companies who need sharing/admin controls.

  1. Confirm plan & workspace. In ChatGPT, check that you’re in your Business workspace (left sidebar).

  2. Review your data policy (share this with staff):

    • “By default, OpenAI does not use Business org data (inputs or outputs) for training.”

  3. Open Admin Console → Settings.

    • Verify org-wide data/privacy controls and who can create or share chats/GPTs. (Business includes admin governance; details evolve as features roll out.)

  4. Teach the team the rule of thumb: “No client secrets in public tools; use the Business workspace only.”


Step-by-Step: If You Use the OpenAI API (builders & automations)

  1. Know the default: API data is not used to train models by default.

  2. Double-check any opt-in: In the platform dashboard docs, confirm you have not enabled data sharing for training.

  3. Rotate keys & restrict scope for contractors and integrations.


Team Playbook (Copy/Paste this into your SOP)

Purpose: Protect client and company IP when using AI.

Allowed tools: ChatGPT Business (company workspace), approved GPTs, and approved API automations.

Rules:

  1. Use the Business workspace for any internal or client work. Never use a personal/free account for sensitive content. (Business data is not used for training.)

  2. If you must use a personal account: Settings → Data Controls → “Improve the model for everyone” OFF.

  3. Do not paste: credentials, unreleased financials, PII, or contract terms into public tools.

  4. Clean up: delete sensitive chats after use; export if you need a record.

  5. API projects: keep the default no-training posture; audit vendors for the same.

One-sentence policy for staff:

“Only use ChatGPT Business for work, and assume anything sensitive stays inside our workspace—never in personal or public tools.”

FAQ (you’ll get these questions)

Q: If I turn off “Improve the model for everyone,” can I still see my chat history?A: Yes—turning that off stops training; it does not hide your history. History visibility is a separate toggle.

Q: We’re on Business. Does OpenAI train on our chats?A: No—by default, Business/Enterprise/Edu data (inputs/outputs) is not used for training.

Q: We used ChatGPT before changing settings. What about old chats?A: You can delete past chats (subject to retention windows) and export your data first if needed.

Q: What about other AI tools (Gemini, Copilot, etc.)?A: Each has its own switch. Review their privacy pages and disable training/data retention where possible. (Here’s an overview comparing toggles across tools.)

5-Minute Hardening Checklist (Main Street Edition)

  •  Everyone uses ChatGPT Business workspace for work.

  •  Personal accounts: Data Controls → “Improve the model for everyone” OFF.

  •  Turn Chat History off on mobile devices handling client data.

  •  Delete sensitive chats after use; export if you need a record.

  •  API projects: confirm no training and vendor parity.


Final word from

Privacy isn’t about paranoia—it’s about process. Choose the right plan, set the right toggles, and teach a one-sentence rule your team won’t forget. Feel free to share this post with your staff and clients.


 
 
 

Comments

bottom of page