Incognito Mode: A Feature Control, Not a Classification Answer

Classification and anonymization answer whether data may be shared with Claude at all. Incognito mode answers a narrower, downstream question: whether a given standalone chat gets remembered afterward. These are different questions, and treating them as the same one is the most common mistake learners make with this feature.

What Incognito Mode Actually Does

Turning on Incognito keeps that particular session from ever entering Memory or your chat history, and it only applies to standalone chats outside of Projects. It's the right toggle for a sensitive conversation or an early-stage draft involving confidential material that you don't want resurfacing later -- floating an idea about a reorg before it's decided, or working through a draft that isn't ready to be tied to your account's visible history.

Critically, Incognito mode does not override the organization's underlying data retention. Turning it on changes what resurfaces to the user later; it does not change what the organization's backend retains or is contractually obligated to retain.

Incognito Does Not Make Regulated Data Safe

This is the trap the exam is built around: flipping on Incognito decides whether a conversation gets retained afterward. It says nothing about whether that data was cleared for this entry point to begin with. For regulated data, the classification question always has to be settled first, and Incognito never stands in for that step.

Worked Contrast: Three Points on the Spectrum

• A confidential M&A document needs summarizing. This is confidential, not regulated -- so after confirming the entry point is policy-approved, Incognito is the right extra step so the conversation doesn't enter Memory or chat history (it still remains subject to the organization's underlying data retention).
• Patient records need summarizing for a healthcare workflow. This is regulated data (protected health information) -- before anything else, you need to know whether this particular entry point is cleared to handle PHI at all. If it isn't, Incognito does nothing to fix that; switching it on doesn't retroactively approve an entry point that was never approved.
• An employee wants to brainstorm icebreaker games for their team's offsite. Nothing here is confidential or regulated, so no governance requirement is pulling Incognito into the decision at all. Flipping it on is a legitimate personal preference -- maybe the employee just doesn't want half-formed brainstorm chatter cluttering their history -- but it's a nice-to-have, not something policy demands. This third point matters as much as the other two: not every privacy decision is a compliance decision. The exam expects you to place a scenario correctly on a spectrum running from 'Incognito can't fix this' (regulated data at an unapproved entry point) through 'Incognito is a sensible extra step' (approved confidential work) to 'Incognito is purely optional' (routine, low-stakes work with no sensitivity at all).

Incognito vs. Project-Scoped Memory Isolation

Both Incognito and Project-scoped Memory isolation (covered in Domain 5) sound like privacy mechanisms, and that surface similarity is exactly why they get confused -- but they solve different problems at different scopes.

A concrete way to keep them apart: Incognito is a switch you flip on a single conversation that shouldn't resurface anywhere later. Project-scoped Memory isolation is a structural property of how Projects are set up -- you don't toggle it per chat, you get it automatically by keeping separate workstreams in separate Projects. If a scenario is about one sensitive chat not showing up in history afterward, that's Incognito. If it's about a fact from one client's Project leaking into a different client's Project, that's Memory scoping, and Incognito would not have prevented it -- Incognito has no notion of which Project, if any, it's running inside.

Common exam traps

• Treating Incognito mode as a substitute for the classification/approval question on regulated data -- it addresses memory and history, not whether the data was allowed here at all.
• Assuming Incognito overrides organizational data retention -- it doesn't; the org's retention policy still applies underneath it.
• Reaching for Incognito as the fix for a policy-prohibited entry point instead of escalating or finding a compliant path.
• Confusing Incognito (single-chat memory/history control) with Project-scoped Memory isolation (cross-Project data separation) -- a scenario testing one is not solved by the mechanism that answers the other.