Stop Oversharing With AI

Before you upload another client report, paste private code, or ask an AI chatbot to “summarize this customer export,” take 5 minutes to check your privacy settings.

The convenience makes it easy to skip this step. A spreadsheet needs summarizing. A customer email needs a response. You attach the file, get an answer, and move on.

But what happens to the information afterward?

Sharing a chat can expose it on the public web, and some shared chats may appear in Google search. Grok says a publicly posted conversation link may be indexed by search engines. In personal ChatGPT accounts, anyone with a shared link can view and forward it, although OpenAI says shared pages are not intended for search indexing. Review the contents before you share. [14, 15]

Depending on the platform, account, and settings, your conversations can remain in your history, become a source for personalization, or be eligible for model training. Safety monitoring and other retention rules can still apply when training is disabled. [1, 5, 8, 10]

For a business, that makes AI privacy part of everyday operations. The information in a prompt might belong to a customer, an employee, or a client who never expected it to be uploaded to a personal chatbot account.

Start with the accounts you actually use. The settings check can be quick; cleaning up old files and reviewing every connection may take longer.

Know which privacy control you are changing

Five separate questions matter:

• Model training: Can the provider use your content to improve its models?
• Storage: How long are chats and files retained?
• Memory: Can information from previous interactions personalize future responses?
• Connected apps: Which external accounts can the assistant access, and what actions can it take?
• Shared links: Who can view conversations you have shared?

Changing one control does not automatically change the others. For example, ChatGPT's training opt-out leaves regular chats in your history, and its app permissions are separate from memory and retention settings. [1, 6]

The steps below focus on personal accounts. Work and school accounts may have administrator-managed controls. Menu names can vary by plan and interface. [1, 5, 7]

01 /Turn off ChatGPT model training

If you do not want new conversations used for model training, open:

Settings > Data controls > Improve the model for everyone > Off

For signed-in users, this preference applies across web and mobile. Your conversations can still appear in your chat history. [1]

Turning off training does not put saved chats on a 30-day deletion schedule. Regular chats remain saved until you delete them or an applicable workspace retention policy removes them. [2]

Also pay attention to feedback. On personal ChatGPT plans, submitting a thumbs-up or thumbs-down can allow the associated conversation to be used for training even after you have opted out. Avoid submitting feedback on conversations that contain sensitive information. [3]

02 /Remove sensitive chats and saved files

Review conversations where you may have shared customer records, internal reports, private documents, or credentials.

In ChatGPT, open the conversation's options menu and select Delete. Archiving keeps the conversation saved.

Deletion removes a chat from your account view immediately. OpenAI generally schedules it for permanent deletion within 30 days, with exceptions for content already de-identified and disconnected from your account, or information retained for security or legal reasons. [2]

Check uploaded files separately. If a document remains saved in ChatGPT Library, deleting the conversation does not delete that file. Remove the saved copy too. We will address memories separately in Step 8. [2]

Also audit shared links. In ChatGPT, go to Settings > Data controls > Shared links > Manage and delete links you no longer need. This stops future access through the link but cannot erase copies another person has already saved. [15]

If you exposed a live API key, password, or access token, revoke or change it promptly. Then update any systems that depend on it and review account activity. Deleting the message cannot invalidate a credential. GitHub's security guidance recommends revoking exposed secrets, generating replacements, and checking activity logs. [12]

03 /Use Temporary Chat for conversations you do not need to keep

For a one-off task in ChatGPT:

1. Open a new chat.
2. Select Temporary.
3. Choose Unpersonalized before sending your first message if you do not want existing memory, custom instructions, or plugins used for personalization.

While the conversation remains temporary, it stays out of your history, creates no new memories, and is not used to improve OpenAI's models. OpenAI may still keep a copy for up to 30 days for safety. If you later save it, it becomes a regular chat and follows your normal settings. [4]

Treat this as a way to reduce persistence. You are still sending information to an external service, so remove unnecessary sensitive details before submitting it.

04 /Review connected apps and permissions

Connecting email, cloud storage, or business tools can give an AI assistant access to more information than you type into a prompt.

In ChatGPT, open Settings > Apps or Plugins, depending on your interface. Review each connected account and disconnect any you no longer need. Where permission controls are available, consider requiring approval before the assistant reads information or makes changes. [6]

For a business account, ask:

• Does this connection support a current task?
• Is it connected to the correct personal or work account?
• Does it need permission to make changes, or is reading enough?
• Who is responsible for reviewing and removing access?

Review third-party privacy policies as well. For example, information sent through a custom GPT action follows the recipient's policies and may be retained longer than a Temporary Chat. [4]

05 /Review Gemini Keep Activity

In Gemini, open Gemini Apps Activity. You can also go directly to Google's Gemini activity page.

Find Keep Activity and choose Turn off. Select Turn off and delete activity if you also want to remove existing account activity. For adult personal-account users, this setting is on by default. [7]

Google says future chats with Keep Activity off are not used to train its AI models unless you submit feedback. It still retains those chats with your account for 72 hours, and safety-related human review can continue. Google also says these settings do not control processing that creates anonymized data to improve its services. [8]

If you prefer to keep activity enabled, choose an auto-delete period. The default is 18 months; three months is an available shorter option. Auto-delete limits stored history but does not itself turn off model training. [7, 8]

Previously reviewed chats may remain for up to three years, disconnected from your account, even after you delete activity. [8]

06 /Review Claude model improvement

For Claude consumer accounts, open:

Settings > Privacy > Help Improve our AI models > Off

Anthropic says this stops new and previously stored chats from being used in future model-training runs. It does not reverse training already started or completed, and separate safety-related uses can still apply. [9]

When model improvement is enabled, Anthropic may retain de-identified conversation data in its training pipelines for up to five years. Turning the setting off does not automatically erase your saved conversations after 30 days. [10]

Delete sensitive chats separately. Anthropic says deleted conversations are removed from backend storage within 30 days, subject to exceptions such as safety and legal requirements. Feedback submissions have separate rules and can make the associated conversation eligible for training. [10, 11]

07 /Use Grok's Private Chat and review training

For a one-off conversation on the standalone Grok website or app, select the ghost icon at the top right to start Private Chat. This keeps the conversation out of your history and excludes it from model training. [14]

Private does not mean immediately erased. Grok's privacy policy says private chats are deleted within 30 days unless retained longer for legal, compliance, or safety reasons. Its FAQ also lists exceptions for de-identified or pseudonymized data disconnected from your account. [14, 16]

To opt out of training for ordinary chats, turn off Improve the Model under Settings > Data on the website, or Settings > Data Controls in the app. Menu labels can vary. The opt-out applies to new conversations; feedback may still be used for training. [14]

Grok inside X has separate controls: Settings > Privacy & Safety > Data sharing and personalization > Grok & Third-party Collaborators. Under Data Sharing, turn off the training permission. Review Grok Personalization separately. [17]

To remove a public Grok share link, visit grok.com/share-links and select Remove. [14]

08 /Review what ChatGPT remembers

Open Settings > Personalization > Memory.

Review the memory summary or saved-memory controls available to your account. Delete sensitive information you do not want used for personalization.

If you do not want information carried across conversations, turn off Memory. Where shown, review Reference saved memories and Reference chat history too.

Remember that deletion happens in separate places:

• Deleting a chat does not necessarily delete a saved memory created from it.
• Turning off Memory does not delete the original chats or files.
• Asking the assistant not to mention something does not delete the underlying source.

Remove sensitive information from the relevant memory controls and the chats or files where it appears. A memory summary is a useful review tool, but it may not display every detail the system can reference. [5]

09 /Share only the information the task needs

Privacy settings work best alongside a simple habit: reduce what you upload in the first place.

Before sending a prompt, ask whether the answer actually requires a real name, exact financial figure, full customer record, or complete codebase.

For many everyday marketing tasks, a smaller, sanitized example is enough to get useful guidance.

For example, you could ask:

Review the sample campaign data below for an ecommerce brand. Identify three reasons conversion rate may have declined and suggest tests to investigate. Client identifiers have been removed, and all figures are fictional.

Use fictional numbers when you need a method or an example. For analysis that requires accurate business figures, use an approved environment and share only the necessary fields. Changing the numbers would change the analysis.

Check the entire attachment before uploading it, including extra spreadsheet tabs, comments, and identifying details. Keep the document that maps placeholders back to real people or clients outside the chatbot.

What should stay out of an ordinary chatbot prompt

Keep passwords, PINs, one-time codes, recovery codes, private keys, and live API tokens out of prompts. Use dedicated credential-management tools for secrets. [12]

For personal consumer AI accounts, make these categories off limits unless an approved workflow specifically permits them:

• Unredacted bank statements, payment details, or identity documents.
• Identifiable customer lists and employee records.
• Confidential contracts, financial reports, or unreleased business plans.
• Private source code and documents you lack permission to share.

If the task requires sensitive information, confirm the approved tool and workflow before proceeding.

Give your team a clear policy for using AI

A privacy audit is more useful when everyone on the team follows the same rules.

Write down which AI tools and account types are approved, what information each may receive, how employees should sanitize inputs, and who to contact after an accidental upload. Make one person responsible for reviewing connections and settings periodically.

Check the actual plan. A paid personal subscription does not automatically provide the same protections as a business workspace. For example, OpenAI says ChatGPT Business, Enterprise, Edu, and its API do not use inputs and outputs for training by default, while personal plans have separate training controls. [3]

Review retention, access, and connected services alongside those training terms. A setting that prevents model training does not answer every question about how business information is handled.

Your AI privacy checklist

• Review optional model training in every personal AI account you use.
• Use native Temporary Chat or Private Chat for conversations you do not need to keep.
• Review shared conversations and remove unnecessary public share links.
• Check sensitive chats, saved files, and memory separately, then remove what should not remain.
• Disconnect unused apps and review the permissions on connections you keep.
• Change exposed passwords and revoke exposed API keys or access tokens.
• Prepare your next prompt with only the necessary information and check the full attachment.
• Confirm your approved tools and account types, and assign an owner for periodic reviews.

Build better habits before your next upload

You can keep using AI to support your work while being more deliberate about what you share.

Start with the training settings in the accounts you use. Remove sensitive chats, files, and shared links, review memory, disconnect unnecessary apps, and prepare a smaller input for your next task.

At Intense City Media, we encourage business owners to approach AI with clear goals and practical processes. Knowing what belongs in a prompt is part of building a workflow your team can use consistently.

Want a clearer plan for using AI across your marketing and business operations? Talk to Intense City Media about your goals, your tools, and where AI can support your growth.

REQUEST A STRATEGY CALL [13]

Official sources

1. OpenAI Data controls in ChatGPT
2. OpenAI Chat and file retention in ChatGPT
3. OpenAI How your data is used to improve model performance
4. OpenAI Temporary chat in ChatGPT
5. OpenAI Memory in ChatGPT
6. OpenAI Managing app permissions in ChatGPT
7. Google Manage and delete your activity in Gemini Apps
8. Google Gemini Apps Privacy Hub
9. Anthropic Model improvement privacy settings
10. Anthropic Consumer data retention
11. Anthropic Consumer model training and feedback
12. GitHub Storing your secrets safely
13. Intense City Media Strategy call
14. Grok Consumer FAQs: Private Chat, training controls, and shared links
15. OpenAI Sharing conversations and scheduled tasks in ChatGPT
16. Grok Privacy Policy: retention of personal information
17. X About Grok: training and personalization settings