Most AI features are turned on by default. Most data collection is opt-out rather than opt-in. And most of the settings that control what happens to your data are buried deep enough that the majority of people never find them.
This is not accidental. The defaults are set in the direction that benefits the companies providing the features. Taking control requires knowing where to look on each platform, because there is no single switch that covers everything.
This guide covers the most widely used AI tools and where to find the settings that actually matter.
Windows 11 and Microsoft Copilot
Windows 11 has several AI data settings scattered across different menus. None of them are grouped together in one place, which means you need to visit each one separately.
Diagnostic Data
Open Settings and go to Privacy and Security, then Diagnostics and Feedback. Turn off Optional diagnostic data. This stops Windows from sending detailed information about how you use your PC to Microsoft for AI model improvement. Basic diagnostic data continues to be sent regardless of this setting, but optional data goes well beyond what Microsoft needs to keep Windows running.
Inking and Typing Personalisation
Still in Privacy and Security, scroll down and find Inking and Typing Personalisation. Turn it off. This stops Windows from collecting samples of what you type, write, and dictate and sending them to Microsoft servers to refine its language models. It is enabled by default and sits quietly in the background without drawing attention to itself.
Advertising ID
Go to Privacy and Security, then General. Turn off Let apps use my advertising ID. This identifier links your behaviour across apps and is used to personalise AI-assisted recommendations and advertising. Turning it off limits cross-app tracking.
Microsoft Recall
If you are on a Copilot Plus PC, Recall takes screenshots of your screen continuously to make your activity searchable later. To disable it, go to Settings, then Privacy and Security, then Recall and Snapshots. Toggle off Save snapshots. Recall does not run at all once this is turned off. For most users, leaving Recall disabled is the appropriate default.
Gaming Copilot
If you use Xbox Game Bar, Microsoft may use your Gaming Copilot chat data to train its AI models. To opt out, press Windows and G to open Game Bar, go to the Gaming Copilot tab, click Settings, and open Privacy settings. Turn off the option that allows Microsoft to use your conversations for model improvement.
Microsoft 365 Apps
Microsoft 365 includes a separate layer of data settings that applies to Word, Excel, PowerPoint, Outlook, and OneNote.
Open any Microsoft 365 application and go to File, then Options, then Trust Center, then Trust Center Settings, then Privacy Options, then Privacy Settings. Uncheck Turn on optional connected experiences. This prevents these applications from analysing your document content and sending it to Microsoft for AI feature improvement.
On Mac, go to the application menu, then Preferences, then Privacy, then Connected Experiences, then Manage Connected Experiences, and uncheck Turn on experiences that analyse your content.
For Microsoft 365 Copilot specifically, the privacy protections depend on your account type. Business and enterprise accounts using licensed Microsoft 365 Copilot have stronger protections by default. Your prompts and the content you share with Copilot are not used to train the underlying models on business accounts. Consumer accounts have fewer automatic protections, and the connected experiences settings above are more important to manage.
ChatGPT
By default, ChatGPT uses your conversations to train OpenAI's models. This applies to free and Plus accounts. Turning it off requires a manual step.
Log into ChatGPT and click your profile picture in the bottom left corner. Go to Settings, then Data Controls. Turn off Improve the model for everyone. With this setting off, new conversations are not used for training.
This setting does not delete your existing conversation history. To do that, go to the same Data Controls section and click Delete all chats. You can also delete individual conversations from the sidebar.
An alternative that does not require changing settings is using Temporary Chat. Click the pencil icon to start a new chat and select the Temporary Chat option. Conversations in this mode are not stored and are not used for training. The trade-off is that the chat disappears after the session ends.
Google Gemini and Google Account AI Settings
Google's AI data settings are spread across your Google Account and the Gemini app specifically.
For your broader Google Account, go to myaccount.google.com and click Data and Privacy, then Web and App Activity. This controls whether Google stores your activity across Search, Gemini, and other services to personalise AI features. You can turn this off entirely or set it to auto-delete after three months or eighteen months.
For Gemini specifically, go to gemini.google.com and click the settings icon. Under Gemini Apps Activity, turn off the toggle. This stops Google from saving your Gemini conversations and using them to improve its models. Existing activity can be reviewed and deleted from the same page.
Google's documentation acknowledges that Gemini interactions may be reviewed by human reviewers unless you disable this activity. Turning off Gemini Apps Activity is the most important step for Gemini privacy.
For Google Workspace accounts such as Gmail, Google Docs, and Google Drive with AI features enabled, data handling is governed by your organisation's Workspace agreement rather than consumer privacy settings. Workspace AI features generally offer stronger privacy protections for business users.
LinkedIn AI Settings
LinkedIn uses your data to train its AI features including writing suggestions, profile recommendations, and feed personalisation. Opting out requires a specific step.
Click your profile picture in the top right, then select Settings and Privacy. Under Data Privacy, find Data for Generative AI Improvement and turn off the slider. This prevents LinkedIn from using your posts, activity, and profile information to train its AI models.
Siri and Apple Intelligence
Apple takes a more privacy-conscious approach by default. Siri requests are processed on-device where possible, and audio is not stored without your explicit consent.
To review your Siri settings on iPhone or Mac, go to Settings, then Siri, then Siri and Dictation History. This shows you the option to delete your Siri history from Apple's servers. Apple Intelligence features run primarily on-device through the NPU and do not send personal content to Apple's servers for processing.
If you use Improve Siri and Dictation, Apple may review a sample of your requests to improve Siri. This is opt-in on recent versions of iOS. Go to Settings, then Privacy and Security, then Analytics and Improvements, then Improve Siri and Dictation to confirm this is turned off if you do not want to participate.
The Default That Matters Most
There is a pattern across every platform in this guide. The default is always set in the direction that benefits the company: data collection on, model training enabled, activity stored indefinitely.
Changing these settings takes about fifteen minutes across all the platforms most people use. The settings do not affect whether the AI features work. They only affect whether your data is retained and used for training after the fact.
For most everyday users, the most important steps are turning off ChatGPT model training, disabling Gemini Apps Activity, turning off Windows inking and typing personalisation, and disabling Microsoft Recall. These four changes address the most significant data collection across the platforms most people use daily, and none of them require any technical knowledge to complete.
Frequently Asked Questions
Does opting out of AI training make the AI less useful for me?
No. Opting out of model training stops your data from being used to improve the AI for future users. It does not affect how the AI responds to you in the current session. The quality of the AI you interact with is determined by the model as it exists now, not by whether your session contributes to future training.
If I delete my chat history on ChatGPT, is it permanently gone?
Deleting your chat history removes it from your account and from use in training for new sessions. OpenAI's data retention policies mean that some data may be retained in logs for safety and legal compliance purposes for a period after deletion. For complete information, OpenAI's privacy policy outlines the specific retention windows that apply to different types of data.
Does using a work or school account change what data AI features collect?
Usually yes, significantly. Enterprise and education accounts for Microsoft 365 Copilot, Google Workspace, and similar services operate under data processing agreements that restrict how the provider uses your data. Your organisation's IT administrator and the agreement between your organisation and the provider govern what is collected and retained. Consumer privacy settings described in this article may not apply or may not be available on managed accounts.
Discussion (0)
Be the first to comment.