
This is a documentation-based review, not a hands-on test. For this October 28, 2026 review, I compared the EmaraChat storefront listing (app ID 6757414605, developer TANSHAU TECHNOLOGY (HK) Co., Limited) with its linked privacy policy and the terms linked from the official homepage, then audited how each promise holds up against the full legal text. I did not install the app, create an account, open a chat, purchase anything, request deletion, or run a security audit. Every statement below is either an attributed developer claim or my reading of published text. Where the record is silent, I say so rather than guess.
The evidence hierarchy used here
For this disclosure decision, I compare the full linked privacy policy, effective January 6, 2026 and last updated September 3, 2026, with the homepage-linked terms, developer-reported App Store privacy label, and promotional description. The privacy label is not verified by Apple. I do not determine which statement would legally prevail: where these sources conflict, a buyer should seek clarification rather than assume the most protective sentence is an unconditional guarantee.
One wrinkle worth resolving early: the store-linked www.emarachat.ai/privacy.html redirects to /privacy, and the apex privacy.html redirects to apex /privacy, with identical complete text. The terms show the same pattern. Hostname and extension differences therefore do not create a second, friendlier operator or a more protective policy. There is one substantive policy, and it must be judged on its own words.
The storefront promise, stated plainly
The listing describes EmaraChat as "designed with privacy at its core," with a "local-first design" where chats "stay private and secure on your device," no data selling "ever," no sharing with third parties, and "robust encryption" protecting conversations. It closes with "your secrets are safe with us." These are strong, categorical assurances that matter when deciding whether to disclose personal material to an AI companion.
Contrast the App Privacy label, which is developer-reported and not verified by Apple: it lists tracking identifiers alongside unlinked identifiers and user content. The tracking-identifiers disclosure complicates the categorical no-sharing promise; it is not proof that data is sold. Selling for money and disclosing information are distinct claims and should be evaluated separately.
Mapping message and device data to servers and providers
The policy's collection section is wide. It says the service may collect account information (email, username, display name, authentication data, preferences), chat and user-generated content (messages, prompts, conversations, images, files, feedback), support messages, and payment and subscription status with transaction identifiers and purchase history. Automatically, it lists IP address and approximate location, device type and model, operating system, app version, language, time zone, network information, unique device and advertising identifiers where permitted, screens viewed and interaction times, crash and diagnostic data, and information gathered through cookies, SDKs, and pixels.
The critical sentence is that chat may reach the company's own servers and infrastructure providers. Section 7 permits provider access where necessary for hosting, storage, authentication, support, analytics, crash reporting, security, payments, communications, and AI infrastructure, plus business transfers, legal and safety needs, and consent. In other words, a conversation composed on a phone may be transmitted to backend systems and specialist vendors. "Local-first" in the description is not defined in the policy as local-only, and nothing in the legal text promises that messages never leave the device.
Section 13 says no identity information is shared with third-party service providers. Read against Section 7, that leaves an unresolved scope conflict, not a clear zero-sharing guarantee. The documents do not explain how these statements fit together, and this review does not decide that one clause legally controls the other.
Consent, training, and the limits of "your data is never used"
Section 4 states that chat content is not used to train or improve the company's own AI systems unless separately authorized, and that a setting may exist. That is conditional, not absolute. It does not promise zero processing of conversations — the service must process them to generate replies — and it does not extend to every vendor's internal practices. Users should not read it as a blanket no-training guarantee across the whole supply chain.
Separately, the policy says consent will be requested where law requires before certain collection or processing. Consent is therefore jurisdiction-dependent and purpose-specific, not a universal opt-in wall. The practical question for a buyer is not "is there a toggle?" but "which processing happens by default, and which requires my affirmative act?" The documents do not fully answer that.
Security language: safeguards, not proof
The policy describes general safeguards and explicitly declines to promise absolute security. That is normal and honest boilerplate, but it means the storefront's "robust encryption" cannot be read as verified end-to-end encryption. No independent audit is cited, and I ran none. Encryption claims here should be treated as unverified marketing until a technical disclosure or audit says otherwise. A breach is not alleged or implied; the point is that the documents do not support the strongest reading of the promise.
Cross-border processing and retention reality
The policy acknowledges possible cross-border processing with appropriate safeguards where law requires. It does not identify specific processing vendors or destination countries, so no specific destination guarantee can be inferred. Retention is described as reasonably necessary for the service, stated purposes, legal obligations, disputes, agreement enforcement, and legitimate interests, with deletion, anonymization, or isolation when no longer needed. Unless otherwise stated in the service, chat history may remain until you delete it or request account deletion, and backups may persist for a limited but unspecified period. That is a realistic description of backend life, and it directly qualifies any assumption that deleting a message removes it everywhere instantly.
Decision table: claim versus policy
| Storefront or label claim | What the full policy actually supports | Buyer reading |
|---|---|---|
| "Local-first"; chats stay on device | Chat may reach own servers and infrastructure providers; vendors named by function only | Not local-only; assume transmission |
| "No data selling — ever" | No sale for money stated; broad disclosure to providers permitted | No-sale ≠ no-disclosure |
| "Never shared with third parties" | Section 7 provider access; Section 13 says no identity info shared | Unresolved scope conflict; request clarification |
| "Robust encryption" | General safeguards; no absolute security; no audit cited | Unverified; not E2E proof |
| App Privacy: tracking identifiers | Advertising identifiers, SDKs, analytics partners listed | Consistent with tracking disclosure |
| Policy restriction on training with chats | Conditional: own systems only, unless authorized; possible setting | Narrow; not zero processing |
Exit checklist before you disclose anything
Because I did not test deletion, treat this as a reading-based checklist, not a verified procedure. First, locate in-app content deletion, which the policy describes conditionally. Second, note that the policy permits an account-deletion request through its contact or the app’s account-deletion function; I did not verify that function. Exceptions apply for law, fraud, abuse, and legal claims. Third, backups may remain for an unspecified limited window. Fourth, expect rights to vary by jurisdiction, with identity verification possible and a legally required response period but no fixed deadline. Fifth, do not assume instantaneous universal erasure or a guaranteed export. If exit certainty is a hard requirement, the documents do not supply it.
Adults only, and what that does not mean
Terms require users to be at least 18, and the privacy policy sets the same floor or a higher local minimum. The terms prohibit pornography, profanity, and sexually gratifying material. Adult eligibility is not permission for uncensored content, and this review is written for adults 18+ only. Separately, store genres, character libraries, world-building, voice, and memory are developer descriptions I did not test; "Only for iPhone" reflects listing compatibility, not proof that no Android app exists. The homepage advertises Android and browser versions, but no package or runtime was verified.
Pricing is secondary and conditional
The US storefront shows Free with in-app purchases, while the description says "always free to chat — no paywalls, no limits." The listing also displays Monthly $9.99, Yearly $39.99, Semiannually $29.99, and named packs plus a Welcome Gift label. I verified no checkout, entitlement, quota, renewal, or refund behavior, so treat "no limits" as a marketing phrase and the prices as displayed, not confirmed. The pack labels do not establish consumability, unit allowances, or unlimited use.
Pros and cons
Pros: a complete, dated policy is published; no-sale-for-money language is explicit; training on the company's own systems is conditioned on authorization; retention and backup limits are acknowledged rather than hidden; adult minimums are stated.
Cons: the storefront's local-first, never-shared, and robust-encryption promises exceed what the policy supports; Section 7 and Section 13 conflict; specific providers, processing-destination countries, and backup durations are unspecified; deletion is conditional and untested; the App Privacy label discloses tracking identifiers.
Buying questions worth asking
Which specific vendors receive chat content, and under what contracts? Is there a default setting that disables training, and is it on or off at signup? What is the maximum backup retention window? Does account deletion reach backups within a stated period? Which identifiers are used for advertising, and can they be reset or refused? Which jurisdiction's rights apply to me, and what is the practical response time? Is the encryption architecture documented anywhere independently? None of these are answered conclusively by the current text.
FAQ
Does EmaraChat keep chats on my device? The storefront says local-first, but the policy permits chat to reach company servers and infrastructure providers. Assume transmission to backend systems.
Is my data sold? The policy denies selling for money. That is narrower than denying all disclosure, and provider access is broad.
Is there a contradiction about sharing? Yes. Section 7 allows provider access; Section 13 says no identity information is shared. Their scope is unresolved; do not assume a zero-sharing guarantee.
Are my chats used for training? The company says it will not use chat content to train or improve its own AI systems unless you separately authorize it. A setting is described conditionally, not verified in the app. This does not establish zero processing or every vendor’s training practices.
Can I delete everything instantly? The policy describes conditional deletion, possible backups for an unspecified period, and legal exceptions. No instant universal erasure is guaranteed.
Is it encrypted end-to-end? The policy cites general safeguards and no absolute security, and no audit is referenced. "Robust encryption" is unverified.
Verdict: proceed without secrets, or wait
For a privacy-and-exit buyer, the decisive fact is that the strongest storefront promises do not survive contact with the linked policy. Local-first is not local-only, no-selling is not no-sharing, and encryption is unverified. The internal contradiction between Sections 7 and 13 means there is no coherent zero-sharing assurance to rely on.
My recommendation is a middle path: you may explore EmaraChat with invented or low-sensitivity material, but do not disclose real secrets, credentials, medical details, or identifying information, and do not treat deletion as guaranteed. If your disclosure is intimate or your exit certainty is non-negotiable, wait until the provider list, backup window, training default, and deletion timeline are documented and independently verified. If you choose a sponsored alternative instead, verify its current terms and privacy policy yourself; we may earn a commission from affiliate links, and a sponsored recommendation implies no privacy or feature parity with EmaraChat.
