Memory is one of the features that makes an AI companion feel continuous. It is also where a private conversation becomes a product data question. A good test uses harmless, fictional details and asks what the controls actually expose.
Map the boundary
Check whether memory is automatic, opt-in or session-only. Look for a visible memory list, an edit action, a delete action and an explanation of whether deleting a memory also deletes the source chat.
Use a safe test
Invent a preference, ask the companion to remember it, and later ask what it knows. Then correct it and test whether the correction persists. Never use a real address, password, health detail, intimate image or another person’s information.
Questions the interface should answer
Who can access conversations? How long are they retained? Is human review possible? Can training use be controlled separately? If the interface cannot answer, label the point unknown instead of guessing.
Pair this guide with the exit checklist and the evidence log.
What we know / what remains uncertain
Known: This article distinguishes observed product behavior, published policy, and expert analysis.
Uncertain: AI products change quickly. Material updates are reflected in the modified date above.
EmberGF links primary sources where available, separates testing from opinion, and labels commercial relationships beside the relevant recommendation.