Aira review: privacy and account control check (com.bytefong.aicompanion)

This review asks a narrow question: if you install Aira on Android under package com.bytefong.aicompanion, what does the developer's own paperwork say about your data and your account, and where does it contradict itself? Aira is listed on Google Play as "Aira: AI Chat & Create" by Bytefong AI Labs, updated 19 October 2026, rated Everyone, with in-app purchases and 5K+ downloads. The two legal documents we examined — a Privacy Policy last updated 2 July 2026 and a User Agreement effective 27 April 2026 — are the evidence base. This is a documentation review, not a product test: we did not install the app, did not buy anything, did not measure performance, and did not verify billing. Everything below is attributed to the developer's published text or flagged as our buying advice.

The privacy question this review is actually built around

Most coverage of companion apps asks whether the characters feel engaging. For a privacy-minded buyer, the sharper question is whether the published record is internally consistent enough to trust with intimate conversation logs. Aira's record is not clean. Three separate mismatches sit in plain sight, and each one matters more than any feature list.

The first is age: the User Agreement says you must be at least 18, and that under-18 use is "strictly prohibited," while the Play listing carries an Everyone maturity rating. The second is sharing: the Play Data safety block declares no data shared with third parties and lists only photos and videos as collected, while the Privacy Policy enumerates account details, profile and personalization data, chat history, optional media, payment status, and device diagnostics, and describes transfers to OpenAI after permission. The third is device data: the User Agreement and the Privacy Policy describe device information differently. None of these is automatically sinister, but each is a documentation gap a buyer should price in.

What the developer says Aira is, and what it is not

Bytefong describes Aira as an entertainment product bundling character chat, image generation, video generation, supported voice interaction, and character discovery. The User Agreement is explicit that AI responses are algorithm-generated "for entertainment purposes only" and that the developer is not responsible for reliance on them. That single clause should reset expectations: this is creative character entertainment, not a guaranteed romance, not a companion with obligations to you, and not a service promising unlimited media output. The Play copy advertises image and video generation as capabilities, but capabilities are not quotas. Nothing in the reviewed documents states message limits, generation limits, or media allowances, so treat any assumption of unlimited use as unverified.

The listing also shows an Apple-centric privacy document. The Policy's payment section says purchases and subscriptions are processed by Apple and describes receiving transaction data "from Apple." For an Android install that is a platform mismatch, not proof that no Android purchase path exists. We flag it as a clarification the developer should make, and we do not invent Android cancellation or refund steps.

Age eligibility versus the Play maturity label

The User Agreement sets a hard 18+ floor, prohibits minor use, and states the developer does not knowingly collect personal information from minors. The Play listing, by contrast, shows an Everyone rating. These can coexist technically — store ratings and contractual eligibility are different instruments — but for a buyer the practical effect is that the storefront gate is looser than the contract gate. If you are evaluating Aira for a household, the contractual 18+ term is the binding one regardless of what the badge suggests. Our advice: treat the Everyone badge as a storefront label, not as permission for a younger user.

The sharing contradiction, stated precisely

We will not soften this into a reassurance. The Play Data safety section, as displayed on the capture date, says "No data shared with third parties" and lists photos and videos as the collected category. The Privacy Policy and User Agreement both describe transmitting chat messages, prompts, and selected character information to OpenAI to generate responses, and both say permission is requested first, with AI features unavailable if you decline. The Policy also describes collecting diagnostics, crash logs, usage events, and support correspondence.

Those statements do not reconcile on their face. A buyer should not conclude "no sharing" from the store badge, and equally should not conclude the transfer is unlawful — permission-gated processing for the core feature is a common architecture. What we can say is that the documentation is inconsistent, that the inconsistency is material for anyone typing personal material into a chat box, and that the developer should reconcile the Data safety declaration with the policy text.

One more boundary: none of the reviewed documents supports claims of no-training, zero-logs, end-to-end encryption, or verified regulatory compliance. The Play block says data is encrypted in transit; that is transport encryption, not end-to-end messaging encryption. The User Agreement says chat content is not sold or shared with OpenAI for advertising and that providers must apply reasonable safeguards. Those are contractual statements, not independently audited facts.

Retention: purpose-based, open-ended, and honest about it

The Policy's retention language is purpose-based and tied to a reasonable period, with no fixed deletion timeline and no stated maximum. That is a real limitation for a privacy buyer, because it means you cannot point to a date by which a deleted conversation is gone from backups or provider systems. The Policy also describes a settings-based history deletion request and a contact channel for personal-data requests. Neither mechanism, as documented, proves that deletion has occurred; they establish a request path. Whether deletion is completed, and how quickly, is not documented with a service level.

Our editorial view: a request-based deletion model is acceptable for a casual entertainment app and weak for an app whose core value is intimate disclosure. If you would be uncomfortable not knowing when a transcript is destroyed, that discomfort should influence your purchase more than any character roster.

Account control and the device-data discrepancy

Account control here rests on credential confidentiality — you are responsible for your password and for activity under your account — and on the deletion request path above. The User Agreement and Privacy Policy both reference device information, but not in matching terms: the Agreement's data language is broader and less itemized than the Policy's diagnostic categories (device ID, app version, language, crash and performance logs, feature usage events, error reports). Buyers should ask the developer to state one authoritative device-data list for Android and to clarify which document governs if they diverge.

This is also where the Apple-centric purchase text compounds the problem: if the Policy's payment and platform language assumes iOS, Android users cannot tell from the documents alone which data flows apply to them. That is a documentation question, not an accusation.

Decision table for the privacy-and-control buyer

Pros and cons on the documented record

Pros. The developer publishes a dated Privacy Policy and User Agreement with a named AI provider and an explicit permission step before chat data leaves the app. The Agreement states chat content is not sold or shared for advertising and requires safeguards from providers. A deletion request path exists in settings and via contact. The entertainment-only framing is unusually direct. The app's stated design intent includes a personal space for conversations.

Cons. The 18+ term conflicts with the Everyone rating. The no-third-party-sharing declaration conflicts with the policy's OpenAI transfer description. Device data is described differently across documents. Payment language is Apple-centric on an Android listing. Retention has no fixed limit and deletion completion is undocumented. In-transit encryption could be misread as end-to-end. No document states message or media allowances. Store reviews, where present, are individual dated testimony and not proof of current pricing or service quality.

Precise questions to put to the developer before paying

Ask in writing and keep the reply. (1) Which document governs if the Play Data safety declaration and the Privacy Policy disagree about third-party sharing? (2) For Android package com.bytefong.aicompanion, what is the complete device-data list, and does the User Agreement or the Privacy Policy control? (3) What exactly does the settings history deletion remove — local cache, server transcript, provider-side copies, backups — and within what period? (4) Is OpenAI processing optional per session, and what degrades if declined? (5) How are Android purchases, renewals, and cancellations handled, given the Apple-centric policy text? (6) Are there message, image, or video quotas at any tier? (7) What is the retention maximum, if any, for chat and media? (8) Will you reconcile the Everyone rating with the 18+ eligibility term?

Hypothetically, a buyer planning to use the app for emotionally sensitive journaling might decide the open-ended retention clause is disqualifying, while a buyer using it purely for light creative prompts might accept it. Both are legitimate readings of the same text; neither is resolved by the marketing copy.

How sponsored alternatives fit — and how to check them

This site is supported in part by labeled sponsorships. Any alternative service mentioned through a sponsor button is a separate product with its own operator, its own policies, and its own eligibility rules. We make no claims about their features, pricing, retention, or encryption, and a sponsorship label is not an endorsement or a comparison finding. Before signing up anywhere, read that provider's current privacy policy and terms directly, check the store's Data safety or privacy label against the policy text the same way we did here, and confirm the age requirement. If a provider cannot answer the deletion-scope question above, that is your answer.

Because the alternatives are independently operated, the honest guidance is procedural: apply the identical contradiction test — age gate versus store rating, sharing declaration versus policy text, retention language versus deletion mechanism — and require a written response before any payment.

Frequently asked questions

Does Aira share my chats with third parties? The Play listing declares no third-party sharing, but the Privacy Policy and User Agreement both describe transmitting chat messages, prompts, and character selections to OpenAI after requesting your permission, with AI features unavailable if you decline. The documents conflict, so the accurate answer is that sharing is described in the policy and denied in the store declaration; we cannot resolve which is operative.

Is Aira end-to-end encrypted? Nothing reviewed supports that. The Play block states data is encrypted in transit, which protects data while moving between your device and servers but does not establish end-to-end encryption where only you and the recipient can read content.

If I delete my chat history, is it gone? The Policy offers a settings-based history deletion request and a contact channel for personal-data requests, but does not state a completion timeline or confirm provider-side and backup deletion. A request mechanism is not proof of erasure; ask for written scope and timing.

Can someone under 18 use Aira? The User Agreement requires users to be 18 or older and prohibits minor use, while the Play listing shows an Everyone rating. The contract term is stricter than the badge; we would treat 18+ as the operative rule and would not rely on the store label.

Does Aira include unlimited image and video creation? No document reviewed states quotas, limits, or unlimited allowances. The listing describes generation capabilities and in-app purchases. Budget on the assumption that generation may be metered, and confirm quotas with the developer before paying.

Is Aira a romance or therapy product? The User Agreement characterizes AI responses as algorithm-generated and for entertainment only, disclaiming responsibility for reliance on them. It is creative character entertainment; it makes no therapeutic claim, and no review should imply one.

The decision

For a privacy-and-account-control buyer, Aira is conditionally usable and not yet trustworthy on paper. The permission step before OpenAI processing and the existence of a deletion request path are genuine positives. The 18+ versus Everyone conflict, the no-sharing declaration versus the policy's transfer description, the divergent device-data language, the Apple-centric payment text, and the open-ended retention clause are genuine negatives that no character feature offsets.

Our actionable recommendation: do not pay for Aira until the developer answers the eight questions above in writing — especially the sharing-reconciliation and deletion-scope questions — and do not treat the Everyone rating, the no-sharing badge, or in-transit encryption as assurances they are not. If you proceed, use it as entertainment with the 18+ rule applied, avoid disclosing sensitive personal material you would not want retained indefinitely, and re-read the Privacy Policy and User Agreement at each update date, since both documents carry version dates that can change the analysis. If the answers do not arrive, apply the same test to any sponsored alternative you explore, on that provider's own current terms.