AI privacy controls and personal data removal

Choose the action that matches the problem

Limit training use. Check the provider’s setting and its scope before sharing personal data. Feedback and other exceptions can have separate rules.

Remove stored content. Delete chats, files, memories or an account through the relevant controls. A training preference alone does not perform these actions.

Report a response. Preserve a specific output and use the provider’s privacy or personal-data request channel when available.

Address the source. If the response cites a public listing, seek removal there as well. Provider controls do not remove that page.

For an ordered checklist, see how to manage personal data across AI services. An AI answer is not proof that your data was in a training dataset.

Find your provider’s controls

Documentation checked October 1, 2026. Signed-in account interfaces were not tested. Settings and eligibility can change.

ChatGPT

Turn off Improve the model for everyone for new eligible conversations, or use OpenAI’s privacy portal. Feedback can still include the associated conversation for training. Chat deletion and personal-data requests are separate.

Read the ChatGPT guide

Claude

Consumer model improvement depends on your saved choice. Feedback, safety review and explicit program opt-ins have separate rules. Commercial products use different terms.

Read the Claude guide

Gemini

With Keep Activity off, future chats are not used to improve Google’s AI models unless you submit feedback. Service and safety processing, including human review, can continue. Deleting activity is separate.

Read the Gemini guide

Perplexity

The consumer AI data retention setting controls training collection after opt-out. It does not delete history or previously collected training data; service and legal processing can continue.

Read the Perplexity guide

Copilot

Choose Microsoft’s documentation for your app version and account. Updated consumer Copilot has separate memory, history, web search and personalization controls. Work and school accounts use different policies.

Read the Copilot guide

Meta AI

Meta documents an EU route to object to public data being used for training. This is a regional process, not a universal account toggle. Check the options for your account, region and type of concern.

Read the Meta AI guide

For X and Grok, use the separate Grok privacy guide.

Where Delist fits

Delist finds exposed personal data, pursues supported removal requests and tracks their status. Review the current scope and any steps that require you. See how Delist works.

The free scan checks public exposure and known breach data within its current scope. It is not an inspection of a provider’s training dataset. A provider may require account access or identity verification for its own controls.

Removing public listings can reduce exposure at those sources. It does not establish a change to a trained model. See how training, retrieval and stored content differ.

Track requests and outcomes separately

Keep the request date, confirmation, stated scope and provider response. A submitted request, an accepted request and a verified removal are different outcomes. Recheck the specific response or source rather than assuming every copy is gone.

Frequently asked questions

Can I opt out of AI training?

Some services offer training preferences for eligible account content. Availability and exceptions depend on the product, account and region. A conversation setting does not necessarily cover public web data or feedback.

Does a training opt-out delete my chats?

Training permission and stored content are separate controls. Check the provider’s chat, file, memory and account deletion options, including retention exceptions.

What if an AI response contains personal data about me?

Keep the prompt, response, date, product and any cited URL. Follow the provider’s personal-data or privacy request process when available. Requests are reviewed under its policy and applicable law; submission does not establish acceptance or removal.

Does removing a broker listing remove it from AI models?

Removing a listing can reduce exposure at that source. It does not prove that a model trained on the listing, undo completed training or remove other copies. Follow up separately on a specific AI response.