CertSafari
    Anthropic Claude Certified Associate – Foundations (CCAO-F) - December 7 onwards· Lessons

    Domain 6 · Lesson 20/21

    Memory, incognito and retention controls for confidential work

    Data Protection in Practice

    9 min read
    5.67% of exam
    6 sources
    Published 18 Nov 2026
    Docs as of 11 Nov 2026

    What you will be able to do

    • Find and delete memory entries that hold confidential client details
    • Choose between an incognito chat, a memory-off chat and a regular chat for a given piece of sensitive work
    • Explain how training use and retention differ between consumer and commercial plans, and for incognito chats, feedback and connector content

    1.What memory keeps, and what deleting a chat leaves behind

    Memory lets confidential details from one conversation reach unrelated work. Claude saves memory as individual topics while you chat. It does not wait to summarize a conversation after it ends. If you mention an unannounced acquisition in a regular chat with memory on, that detail can appear in your next conversation. It can also reach cloud Cowork tasks, because cloud Cowork shares memory with Chat. Each project has its own memory. Claude uses it alongside your account memory, and that is fine as long as both stay within the same engagement.

    Some safeguards are built in. By default, Claude does not store sensitive personal topics such as health, race, ethnicity, religious beliefs or politics, unless you turn on "Include sensitive topics in memory". Some information is never saved even if you ask, including government ID numbers, criminal history, financial account numbers and immigration status. Client business information is not in either of those groups, so a merger codename or deal value can be saved like any other work context.

    Deleting, renaming, closing or starting a new chat leaves memory as it was. Memory entries outlive the conversation they came from. To remove a detail, go to Settings > Your Claude and select Manage next to Your memory. Everything Claude remembers is listed under Topics. Open the topic, then edit it or select Delete, and the change applies to every later conversation. You can also tell Claude in a chat to forget something. That is different from telling it at the start of each chat to disregard what it knows, which leaves the stored entry in place.

    To clear memory more broadly, use Pause or Reset. Pause stops Claude from using memory or making new memories but keeps the existing entries. Reset permanently deletes all memories, including project memories, and can't be undone.

    Checkpoint 1 of 5· Check yourself

    A consultant mentioned a client's unannounced merger codename in regular chats with memory on, and Claude saved it to memory. They want that one detail gone from every later conversation, while keeping the rest of their memory. What should they do?

    Sources1

    2.Incognito chats versus memory-off chats

    For confidential work, decide before the chat starts whether it may build memory. You have two controls, and they differ in what remains afterwards. An incognito chat is not saved to chat history or memory and can't be reopened once closed. A memory-off chat, turned off in the "+" menu, neither uses nor adds to memory. It stays in your history, and searching past chats from other conversations can find it.

    Regular, memory-off and incognito chats compared
    BehaviourRegular chat (memory on)Memory off via "+" menuIncognito chat
    Uses existing memoryYesNoNo
    Adds to memoryYesNoNo
    Kept in chat historyYesYesNo
    Found by searching past chatsYesYesNo
    Available inside a projectYesYes, turns off account and project memoryNo, outside projects only
    When it must be setN/ABefore the first messageWhen starting the chat (ghost icon)

    Two limits decide the choice in practice. The first concerns code. Accounts on the new Claude experience open incognito chats in the previous chat experience, the earlier chat interface, and in that interface Claude can't create files or run code. So an incognito chat on such an account can't analyze a spreadsheet. If you can't confirm which experience your account has, assume incognito can't run code. For spreadsheet analysis that must not touch memory, use a memory-off chat. The second limit is projects: incognito is unavailable inside them, so a memory-off chat is the option there too. Once the first message is sent, the memory setting is locked, and changing it means starting a new chat.

    Incognito chats are not invisible to your organization. On Team and Enterprise plans, they are kept for 30 days for safety or longer under the organization's retention policy. They appear in organizational data exports available to Owners and are included in the Compliance API. Incognito keeps a chat out of your own history and memory. It does not take the chat outside governance.

    Checkpoint 2 of 5· Match them up

    Match each chat setting to how it handles a confidential working session

    Tap a term, then the definition that fits it.

    Checkpoint 3 of 5· Check yourself

    On an Enterprise plan, an analyst discusses a confidential pricing negotiation in an incognito chat and then closes it. Which statement about that chat is accurate?

    Checkpoint 4 of 5· Exam question

    A consultant has memory switched on in their Claude account and has discussed a client's unannounced acquisition in several regular chats. Their firm's policy says client details must not persist beyond the engagement or carry into unrelated work. What is the most appropriate response?

    Sources12

    3.Training use and retention by plan

    Whether a chat is used for training depends on the plan. On consumer plans such as Free, Pro and Max, chats are used to improve Claude if you allow it in Settings > Privacy under "Help improve our AI models". If you turn that off, Anthropic will not use your new chats and coding sessions for future model training. Commercial products such as Claude for Work and the API are not used for training by default. Incognito chats are never used for training, even when Model Improvement is on.

    Thumbs-up and thumbs-down feedback is an exception on every plan. Rating a response sends the entire related conversation to Anthropic, which stores it for up to 5 years and may use it for training. Before you rate a reply in a chat that contains client data, remember that the rating shares the whole conversation. Team and Enterprise Owners can turn off this feedback through the Rate chats setting. Raw content from connectors such as Google Drive is excluded from training and feedback data unless you copy it directly into the conversation. Which connectors to enable is covered under connector management.

    Retention follows its own rules. On consumer plans, a deleted conversation disappears from history immediately and from back-end storage within 30 days. Chats flagged for Usage Policy violations keep their inputs and outputs for up to 2 years and their classification scores for up to 7 years. On Enterprise, when the unified Claude experience is on, a custom retention period that covers chats also applies to memory, except project memory. Project memory is deleted when the project is deleted.

    Checkpoint 5 of 5· Check yourself

    Your company uses Claude for Work, which is not used for training by default. A project manager pastes client contract terms into a chat, likes Claude's summary and clicks thumbs up. What does that rating send to Anthropic?

    Sources34561

    Sources

    (6)

    Every claim above is drawn from one of these pages, quoted as it was written on the date shown.

    1. 1.
      “Claude saves memory as a set of individual topics as you chat, rather than summarizing conversations after they end.”
      ↩︎ What memory keeps, and what deleting a chat leaves behind
      “By default, Claude does not store topics related to personal or sensitive subject matter, like your health, race, ethnicity, religious beliefs, politics”
      ↩︎ What memory keeps, and what deleting a chat leaves behind
      “Some information is never saved to memory, even if you ask. This includes government ID numbers, criminal history, financial account numbers, and immigration status.”
      ↩︎ What memory keeps, and what deleting a chat leaves behind
      “When a conversation is deleted, including under a retention period, memory entries generated from it”
      ↩︎ What memory keeps, and what deleting a chat leaves behind
      “Reset memory: Permanently deletes all memories including project memories.”
      ↩︎ What memory keeps, and what deleting a chat leaves behind
      “Unlike an incognito chat, the chat stays in your chat history and Claude can still find it when searching past chats”
      ↩︎ Incognito chats versus memory-off chats
      “You need to turn memory off before you send the first message.”
      ↩︎ Incognito chats versus memory-off chats
      “it also applies to your memory, except project memory, while the new unified Claude experience is on for your organization”
      ↩︎ Training use and retention by plan
      “Select any topic to read it, then use the edit icon to change it or select "Delete" to remove it.”
      ↩︎ Exam trap 1
      “Select any topic to read it, then use the edit icon to change it or select "Delete" to remove it.”
      ↩︎ Checkpoint
    2. 2.
      “If you have the new Claude experience, incognito chats open in the previous chat experience, so Claude can't create files or run code in them.”
      ↩︎ Incognito chats versus memory-off chats
      “Incognito mode is currently only available for chats outside of projects”
      ↩︎ Incognito chats versus memory-off chats
      “Incognito chats are included in the Compliance API (available for Enterprise plans).”
      ↩︎ Incognito chats versus memory-off chats
      “incognito chats open in the previous chat experience, so Claude can't create files or run code in them”
      ↩︎ Exam trap 2
      “Incognito chats are included in organizational data exports available to account Owners.”
      ↩︎ Exam trap 3
      “Incognito chats are included in organizational data exports available to account Owners.”
      ↩︎ Checkpoint
    3. 3.
      “we will not use any new chats and coding sessions you have with Claude for future model training”
      ↩︎ Training use and retention by plan
    4. 4.
      “Your Incognito chats are not used to improve Claude, even if you have enabled Model Improvement in your Privacy Settings.”
      ↩︎ Training use and retention by plan
      “It does not include raw content from connectors (e.g. Google Drive), including remote and local MCP servers”
      ↩︎ Training use and retention by plan
      “When you provide us feedback via our thumbs up/down button, we will store the entire related conversation”
      ↩︎ Checkpoint
    5. 5.
      “By default, we will not use your inputs or outputs from our commercial products”
      ↩︎ Training use and retention by plan
      “If you explicitly report feedback or bugs to us (e.g. via our thumbs up/down feedback button)”
      ↩︎ Exam trap 4
    6. 6.
      “We retain inputs and outputs for up to 2 years and trust and safety classification scores for up to 7 years”
      ↩︎ Training use and retention by plan

    Ready to test yourself?

    Practise Anthropic Claude Certified Associate – Foundations (CCAO-F) - December 7 onwards in quiz mode.

    Spotted a mistake, or was something unclear? Tell us.