How do I stop my data being used to train AI?

The rules differ by company and by country, and UK and EU users often have stronger options than others. The settings move around, so the reliable approach is to know what to look for rather than to follow a menu path that will be out of date.

Step by step

1. Start with the AI tools you use directly.Look in settings for a section called data controls, privacy, or improve the model. There is usually a switch covering whether chats are used for training, separate from whether history is kept.
2. Note the trade-off before switching it off.On some services, turning off training also turns off conversation history. Decide which you value; on others they are independent.
3. Check the social platforms.Several now use public posts and images to train models. Look in privacy settings for wording about generative AI, model training, or third-party data use. UK and EU users generally have an objection route the platform must provide.
4. Check cloud storage and photo services.Look for settings about improving services, analysing content, or smart features. These are usually separate from the account's advertising settings.
5. Use the business or paid tiers where it matters.Business and enterprise versions of the main AI tools generally exclude customer content from training by default. For anything work-related that is often the practical answer.
6. Exercise your data rights for what already exists.UK data protection law gives you a right to object to processing and to ask what is held. It is a written request to the company, and the Information Commissioner's Office handles complaints if they refuse.
7. Reduce what you put in rather than relying on settings.The durable protection is not typing sensitive information into a chat tool in the first place. Settings change; what you did not share cannot be used.
8. Re-check after major updates.New features frequently arrive switched on. A twice-yearly look at the privacy settings of the services you use most is a reasonable habit.

Tips

• Opting out is not retroactive. A model already trained on data cannot practically have it removed, which is why the timing of the opt-out matters more than its thoroughness.
• Anything you post publicly should be assumed to have been collected, regardless of any platform setting.
• Work accounts are usually governed by the employer's agreement rather than your personal settings. Check the policy rather than the toggle.

Common mistakes

• Assuming deleting the conversation removes it from training — Deletion removes it from your history. Whether it was already used for training is a separate question governed by the setting that was active at the time.
• Relying on a setting instead of judgement — Settings are changed by product updates, sometimes silently. Not entering medical records, financial details or other people's personal information is the protection that does not depend on anyone else.

If it doesn't work

Questions people ask