Tivoraly privacy and Google Play preparation — 29 October 2026
Changes in Preview
- Replaced public Privacy, Child Privacy, Cookies and AI presentation with checked-in, owner-requested factual wording. Added collection/purpose categories, optional child details and uploads, AI request context, location precision, providers, analytics/marketing, access requests and deletion limits.
- Corrected Basic/Pro descriptions, eligible seven-day trial language, complimentary Basic/Pro references, AU/US tax-inclusive advertised totals and existing-subscription currency language. Preserved refund, consumer-rights and other operative billing provisions; no billing change.
- Preserved immediate eligible completion rewards and optional later parent review. Removed hypothetical child messaging/participation controls and complete-export promises.
- Added public
/account-deletioninstructions, support request alternative, footer/Legal Centre links and a Privacy Policy entry in the native parent More menu. Existing signup Privacy links remain. - Separately labelled the 29 October presentation review date. Shared effective versions, acceptance records and legal-review flags remain unchanged; no new effective policy version was installed.
Evidence and limits
Read-only shared queries retrieved current published policies (Privacy/Cookies 1.1, AI 1.2, Child/Membership/Terms 1.3 and AU/US supplements) and retention categories. Every returned retention period is still Requires legal review. Bucket metadata confirms learning-evidence, nearby-photos and worksheet-uploads are private. No private family rows, files or customer billing records were read for this task.
Source review: learner-context.server.ts and prepare.functions.ts attach saved nickname, age/year, interests, learning/support/engagement context to requested AI; AI transports use the Lovable AI Gateway. This does not establish a provider's retention/training agreement. The retired import source file's existence is not evidence of an active advertised feature.
features/analytics/{analytics,config,provider,events}.ts(x) configure GA4, Clarity and Meta Pixel with environment identifiers and consent gates; the root mounts AnalyticsProvider. This materially contradicts stored categorical “no advertising cookies / anonymous-only logging” wording. Live identifiers, provider-side settings, historical collection and a release-specific network inventory were not verified. The revised policy is conditional rather than asserting all three run in Live.
Nearby location.server.ts processes typed Google Places queries and rounds area centres to three decimals; features/nearby/store.ts stores preferences locally. The Android manifest has camera/audio permissions, not coarse/fine location; no operative GPS lookup was found in this review. Voluntary addresses, typed areas and network metadata are not equated to absence of all location processing.
account.functions.ts schedules a 30-day restoration deadline after cancelling billing. deletion-cleanup.server.ts removes owned files before erasure and can hold failed cleanup for staff. purge-launch.server.ts remains false/inactive. No new claim of active purge suspension or exact-time deletion was made. External workers/privileged paths and deployed server equivalence were not re-certified.
This is current source plus shared policy/configuration evidence, not certification that all Preview changes are already Live. No paid requests, customer writes, policy-store changes, migrations, games/account changes or publication.
Google Play requirements still unresolved
- Owner/legal entity and contact: Confirm the legal operator identity displayed in policy/store listing and privacy contact process. Existing [email protected] is reused, not invented.
- Retention and providers: Approve actual category periods, backup erasure, provider/subprocessor names/locations, AI retention/training/human access, email retention and request handling. No invented schedule or no-training promise.
- Children/Families: Confirm actual target ages and direct Child Mode use; obtain qualified AU/US child privacy/COPPA consent review. Audit child/unknown-age analytics/SDK identifiers and Clarity masking. The current analytics loader has no child-age gate: parent consent and event filtering alone do not establish Families compliance. This task changes wording only, not tracking behaviour.
- Data Safety form: Map the shipped Android build and backend providers for parent identity, optional child profile/support details, learning/user content, photos/video/audio/files, approximate/location-related data, account/payment identifiers, device IDs and diagnostics. Determine required versus optional collection, sharing/service-provider exceptions and purposes using Google's definitions. Confirm encryption-in-transit operationally before answering; HTTPS URLs alone are not a complete attestation.
- Account deletion: Publish an approved public deletion resource and verify the in-app request/support process on the actual Android build. Confirm final deletion timing/exceptions and operator worker coverage. This Preview URL is not yet a publicly released Play deletion URL.
- Android sensitive permissions: Confirm camera/microphone disclosure and runtime requests match actual capture/voice workflows and any prominent consent requirements. Manifest evidence is not a physical-device check.
- AI reporting: Assess whether AI-Generated Content rules apply and whether the shipped app provides the required in-app reporting/feedback for covered generated content; a support link is not a verified substitute.
- Payments: Verify Play release never sells or steers to Stripe digital-membership checkout outside an applicable enrolled exception. No native purchase flow was changed or certified here.
- Release consistency: Approve final policy version/effective date, notices/reacceptance and publication separately; confirm Play Console links and Data Safety answers match the exact app/website release. Existing review flags remain intentional, not a claim of readiness.
Official Google references
- User Data / privacy policy: https://support.google.com/googleplay/android-developer/answer/10144311
- Data Safety: https://support.google.com/googleplay/android-developer/answer/10787469
- Account deletion: https://support.google.com/googleplay/android-developer/answer/13327111
- Families: https://support.google.com/googleplay/android-developer/answer/9893335
- Target audience: https://support.google.com/googleplay/android-developer/answer/9867159
- AI-generated content: https://support.google.com/googleplay/android-developer/answer/14094294
- Payments: https://support.google.com/googleplay/android-developer/answer/10281818
Follow-up implementation audit — 29 October 2026
Assessment: not ready to claim the remaining privacy issues are resolved. This focused audit used source, existing read-only evidence and official provider documentation. It did not inspect customer records, access vendor account settings, run deletion/payment/provider operations or verify a shipped Android binary. “Source verified” below means the inspected application implements that behaviour, not that the published release or every privileged pathway matches it.
Analytics and withdrawing consent
- Source verified: GA4/Clarity load after analytics consent; Meta Pixel after marketing consent. Missing identifiers prevent loading. Default consent denies both; choices/version/timestamp are stored locally and expire after 365 days. GA4 receives page path/title/location and allowed product-event fields; Meta receives page views and mapped registration/trial/payment/search events. Vendor scripts can additionally process network/device/cookie/interaction data. Clarity supports session recordings. No vendor account settings or actual historical collection were verified.
- Evidence:
src/features/analytics/analytics.ts:45–146,consent.ts:20–85,events.ts:1–84,provider.tsx:17–36. GA settings disable signals/ad personalisation, but do not prove anonymity. Some DOM masking and a comment recommending project-wide masking are not proof of the provider's masking configuration. - Policy accuracy: vendor/category disclosures were already conditional and accurate. Corrected withdrawal wording:
applyConsentonly loads permitted vendors; it does not revoke vendor consent or unload existing scripts. Consent-gated application events stop, but vendor collection in the current page may continue. Added page/title transmission and local choice expiry. - Code versus wording: stopping already-running vendor collection reliably requires a separately approved code change and vendor-specific validation; wording alone does not fix it. A child/unknown-age exclusion is also absent. No tracking code was changed.
- Decision/confirmation: decide which optional vendors to retain, and whether tracking should be excluded from all signed-in family/Child Mode surfaces. Confirm enabled production identifiers, project-wide masking, retention and deletion settings. Do not use this audit as evidence that optional trackers are enabled in Live.
Android tracking and permissions
- Source verified:
capacitor.config.ts:13–27bundles screens locally and has no remoteserver.url. The web tracker environment check permits production domains or explicit debug mode (src/features/analytics/config.ts:14–40); it does not explicitly distinguish native use. Reviewed direct dependencies contain no dedicated analytics SDK. The checked-in Android manifest declares network, camera and microphone permissions, not coarse/fine location. Google Services is conditional inandroid/app/build.gradle:48–53. - Data: website-style tracking could occur in allowed hosted pages or debug-enabled contexts. Capture/voice features can process selected photos/audio; manifest permissions alone do not prove collection or runtime permission prompts. Backend processing still occurs regardless of whether optional web tracking loads.
- Policy accuracy: added these qualified source facts, not a “no Android tracking” claim. Final transitive SDKs, merged manifest, bundled settings, installed-app traffic and Play Data Safety answers remain unverified.
- Code versus wording: a release-specific Android network/permission review is needed before deciding whether changes to native tracking or permission disclosures are required. No Android code was changed.
- Decision/confirmation: confirm intended child age audience and provide the exact release for verification; decide the native optional-tracking policy. Source-only inspection cannot certify a Play release.
Children, parent authority and notice acknowledgement
- Source verified: email signup combines Privacy/Child Privacy acknowledgement and separately requires Terms (
src/routes/sign-up.tsx:110–116,343–367).src/features/legal/acceptance.ts:17–33,39–53records account/document version, acknowledgement details, country/locale, source page and browser user-agent where supplied; parent responsibility records are separate. This is not verified parental authority or age verification. Acceptance failure handling and social-signup equivalence were not certified in this audit. - Data: saved profile write/read paths include interest strength, parent/support/avoid notes, flow supports/notes and structured engagement preferences (
src/lib/learners/learning-profile.functions.ts:20–45,77–98).learner-context.server.ts:43–98reads these plus confirmed preferences and builds AI context. Learning records, optional uploads, rewards and child-facing assistance may add further content. Support/engagement text can reveal sensitive information. - Policy accuracy: expanded the optional-child-data inventory and acknowledgement metadata. Replaced the absolute assertion that children “cannot independently sign up” with the adult-account requirement and explicit lack of independent age/authority verification. Child profiles still have no separate sign-in. Added the concrete absence of a Child Mode analytics exclusion; parent cookie consent can persist into child use.
- Code versus wording: consent/age/authority controls and child tracking restrictions need a separately approved implementation if qualified review requires them. Do not describe a checkbox, PIN or payment card as sufficient legal parental consent. Existing access controls and rewards were untouched.
- Decision/confirmation: confirm the intended ages, obtain qualified AU/US child-privacy/consent advice, and decide what optional sensitive context families should be invited to provide and send to AI. No compliance conclusion was reached.
Third-party processing, retention and backups
- Source verified: AI transports target the Lovable AI Gateway (for example
src/lib/ai/responses.server.ts:22); saved child context is assembled by the app. Firecrawl/Places receive feature-specific search/location requests; app messages use the managed sender (src/lib/email-templates/send-email.ts:35–69). Auth emails are a separate path. These source facts do not identify the complete downstream contractual processor chain or deployed settings. - Data: prompts/context, selected report/activity content and import/voice material where offered; search queries and typed places; recipient/address and message content for email; account/learning/files for cloud services. Application usage logs and opted-in AI result reuse can also retain outputs separately from external providers (
src/lib/ai-usage/limits.server.ts). - Policy accuracy: existing explicit unknowns remain accurate; no no-training, no-retention, fixed-region or complete-provider-erasure promise was added. Cloud documentation describes daily database-only backups retained up to roughly 14 days, excluding storage files; that is generic service documentation, not this project's verified backup inventory or deletion commitment.
- Provider evidence: Lovable DPA and subprocessors need confirmation of applicable contract/product scope. Firecrawl security describes enterprise/opt-in zero-retention controls, not a verified setting for Tivoraly. Google Places policies describe API use rules, not this project's retention agreement. Direct OpenAI/Anthropic API terms must not be assumed to govern a Lovable-mediated request.
- Code versus wording: provider settings/contract clarification and an approved retention schedule come first; minimisation/redaction or automated retention enforcement may then need code changes. This audit did not change transport, caching, model selection or retention behaviour.
- Decision/confirmation: confirm legal operator, applicable provider agreements/subprocessors/regions, training use, prompt/log/email retention, human access and deletion terms. Approve real category-specific retention periods only after confirming they can be honoured.
Account deletion, restoration and retained data
- Source verified: grown-up/email-confirmed requests cancel billing before setting a 30-day deadline (
src/lib/account/account.functions.ts:19–67). Restoration uses the existing pending-deletion flow; source is not proof of full deployed race/deadline enforcement. The cleanup worker holds on failed/uncertain file removal and attempts files before final erasure (deletion-cleanup.server.ts:44–145). The Preview suspension flag is false (purge-launch.server.ts:2). Prepared pause text is an unused export (deletion-notice.ts:1–2), not evidence of an active pause./account-deletionimports its instructions directly and invokes no deletion action. - Data: pending accounts/files remain until removal; published/shared material, operational records and one-way trial-use identifiers may receive different treatment. Hashing is not proof of anonymity. Provider copies/backups are not erased by assumption.
- Policy accuracy: qualified the file-first promise as the application worker's design, not all privileged/scheduled pathways. Retained the 30-day window, no guaranteed erasure timestamp, no automatic billing restart and no active-pause announcement. Existing evidence identifies independent privileged/database pathways; current operator coverage/drain remains unconfirmed.
- Code versus wording: coordinated operator/pathway controls and any required restoration-race fix need separate approval. Publishing pause wording without a proven coordinated pause is not safe. No deletion, worker, database or customer data changed.
- Decision/confirmation: confirm the responsible operator and complete pathway coverage/drain before any purge-status announcement; approve actual retention exceptions and request handling. The 10 November review date is a proposal, not an automatic restart or currently activated suspension.
Data access and export
- Source verified: reports are selected editable snapshots, with learning/progress/portfolio/activity/custom types (
src/features/reports/types.ts:2–15,sources.ts:161,380–400). Available report/document downloads are not an all-account export;src/routes/trust.tsx:167–172already says this. A project-owner database export is not a family-facing export and can include other families; it must never be supplied as an individual privacy response. - Data: downloads contain selected learning records/documents, not necessarily all profiles, preferences, authentication/billing/operational records or original files. The support link requests handling; it does not implement or verify a complete fulfilment workflow.
- Policy accuracy: the existing “not a complete account-data export” and support-request wording is accurate. No new export capability, completeness, response deadline or format is promised.
- Code versus wording: decide and document a secure, identity-verified, family-scoped manual fulfilment process. A self-service complete export is a separate feature requiring approval, not necessary to pretend existing downloads are complete.
- Decision/confirmation: choose who handles requests, which data/files can be supplied, secure delivery and applicable response deadlines; obtain qualified advice on rights/exceptions.
Focused verification for this follow-up
Three existing network-blocked legal boundary tests passed; stored input/metadata and unrelated documents remain unchanged. Signed-out Chromium read Privacy, Cookies, Child Privacy, account-deletion and legal-review pages: all HTTP 200 with the corrected disclosures present in the rendered main content and zero page errors. The initial network-idle wait timed out; repeating with document/visible-text waits passed. Cookie screenshots still showed a loading view despite the successful text checks, so visual screenshot verification is inconclusive and is not claimed. Preview compilation reported build OK. No tracking scripts were deliberately enabled, no installed Android test was performed, and no provider settings, customer records, shared legal records, account/billing/game functionality or publication were changed.
Verification
Three focused presentation-boundary tests passed: stored input/acceptance metadata unchanged; unrelated policy and unrelated country supplement preserved. Signed-out Chromium opened Privacy, Child Privacy, Cookies, AI, Membership, Terms and account-deletion pages (all HTTP 200), clicked Privacy → deletion and homepage footer → Privacy, with zero page errors. Legal/deletion pages are now exempt from the country-selection overlay, so direct privacy requests are readable immediately. Screenshots of the Cookie and deletion pages were inspected. Preview compilation reports build OK.
Android parent More now contains a dedicated Privacy Policy link; signup links were already present. No signed-in session was created and no installed Android/physical-device launch or Play Console link check was performed. No full account deletion, real subscription or native release test is authorised or claimed.