AI memory privacy, with controls you can use

History, context, memory and cache do different jobs

Saved history is the conversation you can reopen. Context is the material supplied for a particular reply: messages, relevant files, instructions and other selected information. Memory carries useful information between conversations, such as your writing preferences or a continuing project's decisions. Cached input tokens refer to repeated model input that a provider can reuse during processing; they are not a separate personal memory you edit. These categories can overlap. A preference may appear in a saved chat, the memory document and a later request's context. Clearing one does not mean every other copy is removed. Start by identifying which item you want to change: the conversation record, a saved preference or the material used for today's answer.

Review what Quad remembers

In Memory settings, you can read the saved memory document, edit its contents, save your changes or clear it. Quad also adds useful durable information as you chat, so memory is not limited to text you entered into that editor yourself. You can ask the assistant to remember a preference, correct a detail or forget something saved there. Replace an outdated preference instead of leaving contradictory instructions. Remove project details that no longer help future replies. Review the relevant saved conversation separately if you also want to remove its record. The private workspace guide explains how chats, files and projects fit together. Memory settings control the saved document, not the entire account's history.

Sensitive memory is a separate choice

Sensitive memory is off by default. Privacy settings provide a separate optional permission for storing sensitive details for future replies. When enabled, the service is instructed to save a sensitive fact only when you explicitly ask it to remember that exact detail. It is not permission to infer protected traits or collect everything mentioned in a conversation. Remembered details used in future replies are shared with the providers processing those replies. Disabling sensitive memory stops new sensitive memory extraction; it does not delete existing memories or reverse earlier transmissions. Review and remove saved details in Memory settings as well. The privacy policy explains this permission and the separate permission for third-party AI processing.

Use Incognito for a conversation that stays separate

An Incognito conversation does not read or update your memory and is not saved to your chat history. Temporary chat data is deleted from Quad's servers within an hour. This is useful for a question that should not shape later assistance. It does not remove memories already saved by earlier conversations, and it does not stop the selected model from processing the current request. For a normal saved conversation, review the information you attach and the instructions you provide. Context can change between requests, so an answer's input size is not a count of permanent memories. See the private chat guide for how history and Incognito work.

Check provider retention and sharing separately

Quad's saved memory and model provider retention are separate. Private Lane chat routes use Zero Data Retention agreements. Standard Lane models require confirmation and follow the provider's standard retention terms. A cached-token count describes processing reuse, not a privacy promise or proof that a provider keeps nothing. The retention guide explains the lanes and their boundaries. Quad does not use your conversations or files to train its own models. Saved account content is encrypted in transit and at rest. Sharing a conversation or publishing an output gives others access to the content you distribute; review that choice before sending a link. The security page covers protections, while your memory settings let you keep future assistance relevant.

Frequently asked questions

Is AI memory the same as chat history?

No. Chat history is the saved conversation record. Memory is useful information carried into later conversations. A fact can appear in both, so review the memory document and the original chat separately when changing or removing it.

How do I edit or clear Quad's memory?

Open Memory settings to read and edit the saved memory document, save changes or clear it. Quad also updates memory as you chat. Editing the document does not by itself remove the original saved conversation.

Does Incognito use my saved memory?

No. Incognito conversations do not read or update memory and are not saved to chat history. Temporary chat data is deleted from Quad's servers within an hour. The model provider still processes the request to answer it.

Does disabling sensitive memory delete existing details?

No. Disabling the optional permission stops new sensitive memory extraction. Review and delete existing details in Memory settings separately. It does not undo previous provider transmissions or delete saved chats.

Are cached input tokens saved memories?

No. Cached tokens describe repeated input a model provider can reuse during processing. They are different from the memory document you can edit in Quad. Cache usage does not replace checking the chosen provider's retention terms.