AI companion apps now vary from simple character chat to real-time voice, video and photo-based digital humans. Before giving any app access to your face, microphone or personal history, review the difference between operating-system permissions and the service’s own data practices. This practical checklist works for iPhone and Android without pretending that every product uses the same settings or privacy policy.

Start with the app listing, not its marketing claims

An AI companion app can offer conversations, voice notes, interactive video, custom characters and photo-based avatars. Those features may legitimately require different kinds of information, so ‘private’ should never be treated as a single feature that is either present or absent. Before installing an unfamiliar service, identify which parts of the app you intend to use. A person who wants only text chat has a different risk profile from someone uploading their face, voice samples and personal conversations.

On Android, Google Play’s Data safety section describes information the developer declares about collection, sharing and handling. Google explicitly distinguishes these disclosures from the technical permission list. The two are useful together, but neither should be mistaken for a complete independent security audit. On iPhone, review the App Store privacy information and the developer’s own policy. Check for a real operator, contact route, clear account deletion instructions and explanations of data used to personalize the experience.

Check permissions one by one

Camera and microphone access are obvious in a video or voice companion, but that does not mean they should be enabled continuously. Apple’s official instructions show how iPhone users can inspect permissions in Settings under Privacy & Security and change access to hardware features. Android’s help pages similarly describe the Settings, Apps and Permissions path. Exact labels can vary by device maker and software version, so follow the device’s current interface rather than assuming screenshots from an older guide still apply.

Think in terms of the minimum permission required for each activity. Text conversations may not require a camera. Choosing a single image from a picker may differ from giving broad photo-library access. When a feature is not in use, consider revoking unnecessary access and see whether the rest of the experience still works. A permission denial may disable one function without making the entire product unusable. Do not assume, however, that turning off a device permission removes files previously uploaded to a service.

Understand what leaves the device

Device permissions govern access to hardware and local data, while a privacy policy addresses how the service processes information after receiving it. These layers are different. A text prompt, uploaded face, voice recording or generated image may be processed on remote servers even when the app no longer has access to the microphone or camera. Review whether content is used to deliver a feature, personalize an account, improve models, or shared with contractors, and whether users can change those settings.

Pay special attention to retention and deletion. Some providers separate chat-history deletion, account deletion and withdrawal of training consent. If a product says it supports exports, find out which data categories are included and whether the resulting file is readable in another service. Keep a copy of the policies you relied on if you plan to upload highly personal material. Product settings can change between versions; it is reasonable to review them after a major update.

Test the privacy experience with harmless content

Before uploading private photos or revealing sensitive personal details, use a fictional test conversation and a non-sensitive image you have the right to share. Check which permissions are requested, whether your chosen image appears in the public profile, and whether you can remove the test asset. Explore the profile’s visibility, character-sharing options and message reporting tools. A service with easy creation but no discoverable removal workflow deserves closer scrutiny.

Also inspect the session on a second device if you normally switch between phone and tablet. Determine whether chat history syncs, what remains accessible after logging out, and whether there is an option to sign out of other sessions. Avoid treating a convenient ‘clear chat’ control as proof that all server data has been erased. The goal of the test is to understand available controls, not to make unsupported claims about a company’s internal infrastructure.

Compare iPhone and Android privacy tools realistically

iPhone provides camera and microphone indicators, while Android offers permission management and, on supported versions, a privacy dashboard that shows recent use of sensitive permissions. These features help identify when device capabilities were accessed. They do not by themselves reveal the full content of an app’s server-side processing, whether an AI model memorized a message, or every downstream data-sharing arrangement. This distinction protects users from an inaccurate feeling of certainty.

On a shared device, check notification previews as well as access permissions. A companion’s character name, message excerpt or image thumbnail may appear before anyone opens the app. Consider whether notifications should display on the lock screen and whether the app supports its own screen lock or account separation. These controls vary by product; do not assume every AI companion includes them. Review the specific features you can actually find and test.

A final decision checklist before paying

Make the purchase decision only after answering practical questions: Who operates the app? What does it collect for text, voice, photos and video? Can you control profile visibility? Does the app explain retention and account deletion? Can unnecessary permissions stay off? Are the subscription terms, renewal price and credit charges clear? Is there a direct route to support or report misuse? A polished avatar demo does not answer any of these questions.

No guide can certify the privacy of every AI companion without testing and auditing each version. Instead, compare verifiable controls and the quality of the provider’s explanations. If you cannot find clear rules for deleting your face or voice, avoid uploading those materials until you receive an adequate answer. A satisfying AI companion experience should not require surrendering more personal information than the feature actually needs.

Official references: Apple Support: iPhone hardware permissions; Android Help: Change app permissions; Google Play: Data safety section; Android privacy dashboard.