Home/AI companion privacy & safetyKeep your boundaries close.
AI companion privacy & safety
Understand what you share, what a service may retain, and which controls are actually available. Privacy starts before the first intimate conversation.
What we check before we recommend a privacy control.
Safety pages on EmberGF are not a dump of vendor claims. Each scorecard starts with a live account: we create it, change visibility and training toggles where they exist, request an export, and follow the published deletion path. We record whether those controls are actually present, whether they persist after a refresh, and whether the help centre contradicts the in-product labels.
We also separate conversation tone from data handling. A companion can feel private while still logging prompts, billing identity, or device metadata. Our notes cover account email requirements, payment descriptors, session cookies, and any statement about model training.
Use the cards below as a reading list, not a checkout. One labelled partner link may appear after this explainer for readers who already finished a review. Editorial links stay first-party; sponsored routes use /go/ and rel="sponsored nofollow". Adults only.
Editorial note: This guide is about verifying the controls visible to you today. Product interfaces, plans and policies change, so check the live account…
Editorial note: This guide is about verifying the controls visible to you today. Product interfaces, plans and policies change, so check the live account…
Editorial note: This guide is about verifying the controls visible to you today. Product interfaces, plans and policies change, so check the live account…
Editorial note: This guide is about verifying the controls visible to you today. Product interfaces, plans and policies change, so check the live account…
Separate a private-looking chat from a private data policy. Look for visibility controls, training settings, deletion instructions and retention exceptions. Use fictional details while testing; a reassuring reply is not a guarantee about how a service stores information.