Trace your commute

AI conversations and connected data

What AI chats can retain beyond the conversation list

Provider examples show how AI-chat history, model improvement, feedback, temporary sessions and connected services produce different records.

Sources checked .

Conceptual illustration

Consumer training choices have specific exceptions

Anthropic's consumer guidance covers Claude Free, Pro and Max, including Claude Code used with those accounts. It describes model improvement when users allow it, separate explicit opt-ins, and safety-review uses. Its scope differs from commercial products. The setting should not be read as a promise that no provider processing occurs. Anthropic: consumer training choices, safety review and feedback

Removing a chat from view is not instantaneous backend deletion

Anthropic says a deleted consumer conversation disappears from history immediately and is deleted from backend storage within 30 days. It separately describes up to five years for de-identified model-training data when improvement is allowed, and longer retention for specified safety, legal or dispute purposes. Turning off improvement does not undo training already underway or completed. Anthropic: consumer deletion, training retention and exceptions

An incognito label does not mean no retained record

Claude's help says incognito chats are absent from history and memory and are not used for training. It describes 30-day default retention, with longer organizational retention possible. Team and Enterprise incognito chats can be included in owner exports, with Enterprise Compliance API coverage. This mode is not secrecy from every organization administrator. Anthropic: incognito history, retention and organization exports

Gemini has several separate retention paths

Google says temporary chats are not used for model training, but safety processing, including human review, can still occur. Previously reviewed chats can remain up to three years after activity deletion. Google: Gemini Apps privacy and retention

The visible history and public links have different controls

Google's personal-account activity guide says activity-off chats remain with the account for up to 72 hours. It provides review, deletion and auto-delete choices. It says deleting a shared chat does not delete its related public link, and deleting a conversation with a Gem does not delete the Gem. Work or school activity settings can instead be controlled by the organization's administrator. Google: personal Gemini activity controls and separate public links

Temporary Gemini chats have availability limits

Google documents temporary chats for signed-in personal accounts on its web or mobile app. They are not available through work or school accounts, Gemini in Chrome or Gemini in Google Messages. These chats do not appear in recent chats, and access is lost after leaving them. Google: temporary-chat availability and feature limits

Connected information can become part of an AI interaction

For eligible Personal Intelligence users, Google describes summaries, excerpts and inferences from connected data being used for responses and model improvement when Keep Activity is on. It distinguishes that from training directly on entire source inboxes or libraries. Disconnecting an app does not delete Gemini activity; deleting Gemini activity does not delete the other service's records. Google: Personal Intelligence connected-app data and deletion scope

Feedback can have its own record

Anthropic says consumer thumbs-up or thumbs-down feedback can retain the related conversation and included content for up to five years. Its training guidance distinguishes raw connector content from content copied into a conversation. Sending a feedback report is a separate data decision from simply reading an answer. Anthropic: consumer training choices, safety review and feedback

Enterprise commitments must keep their plan scope

OpenAI's ChatGPT Work admin FAQ describes no training on ChatGPT Enterprise business data by default, alongside workspace controls. It also says Work with local access does not provide strict zero retention. Connected services keep their own retention and access requirements. Those enterprise statements should not be applied as blanket rules for personal ChatGPT accounts. OpenAI: ChatGPT Work enterprise data commitments and connected services

Questions and answers

Does an AI-chat training opt-out mean no retention?

Anthropic documents retention and legal or safety exceptions separately from model improvement. Anthropic: consumer deletion, training retention and exceptions

Does deleting a Gemini chat remove its public link?

Google says the related public link is managed separately. Google: personal Gemini activity controls and separate public links

Can a work organization retain an incognito Claude chat?

Claude documents organization exports and potentially longer Enterprise retention. Anthropic: incognito history, retention and organization exports

Sources and scope

Public provider documentation checked 10-24-2026. Examples cover personal Gemini Apps, Claude consumer controls and specified enterprise features. OpenAI examples are limited to the cited ChatGPT Work enterprise documentation.

Keep exploring