Privacy Policy

This Privacy Policy explains how Copycat Cafe Limited (“Copycat Cafe”, “we”, “us”) collects, uses, shares, and protects personal data when you use our website, app, language-learning features, subscriptions, emails, support, and related services.

1. Who controls your data

2. Data we collect

• Account and invitation data: name, email address, login details, account settings, language preferences, communication preferences, account-invitation recipient email, and Copy Pass details such as sender, link status, claimant account, and promotional access dates. Copycat Cafe does not ask for or store a Copy Pass recipient’s email before they claim the link.
• Subscription and billing data: subscription plan, status, renewal, cancellation, refund, transaction, Paddle customer/subscription identifiers, and limited billing metadata. We do not store full card details.
• Learning data: lessons watched, exercises completed, answers, translations, flashcards, progress, level, language choices, chat messages, corrections, scores, and feedback.
• Voice, audio, and speech data: recordings you submit, live microphone audio while you use speech features, transcripts, pronunciation scores, target phrases, generated audio, and related provider metadata.
• Recording diagnostics: bounded measurements and categories such as recording duration, volume, clipping, silence, estimated background noise, browser/operating-system/device category, audio format, and microphone-track settings. These diagnostics do not include the recording itself, a raw user-agent string, or a persistent device identifier.
• AI data: prompts, chat context, lesson context, model outputs, evaluator results, and metadata needed to generate responses or assess practice sessions.
• Support and contact data: messages you send us, testimonial submissions, support history, and information needed to answer you.
• Technical, security, and usage data: IP address, user agent, device/browser data, logs, error reports, cookies, session identifiers, form-security signals, analytics events, and referrer/affiliate information.

Public share links: when you confirm that you want to create a public result link, the page displays the information identified before publication, such as the public display name shown there (your first name by default, or “A learner” if you remove it), your result, copied phrase or conversation highlight, language, and any included share image. Anyone with the link may view, reshare, screenshot, or cache it. The page does not publish your raw pronunciation recording or private chat transcript. New links expire after 30 days. You can change the public name, revoke a link immediately, or delete the result and its share image from Privacy & sharing in your account settings. These controls cannot remove copies already saved, reposted, screenshotted, or cached elsewhere.

Special-category and highly sensitive data

Copycat Cafe is not designed to collect special-category data such as health information, political opinions, religious beliefs, sexual orientation, biometric identifiers, or similar highly sensitive information. Please do not include highly sensitive personal data in practice conversations, recordings, support messages, or screenshots unless it is necessary for your request. If you choose to include it, we process it only as needed to provide the service, respond to you, maintain security, debug issues, comply with law, or protect legal claims.

3. Why we use data and our legal bases

• To provide the service under our contract with you: accounts, lessons, AI chat, speech recognition, pronunciation scoring, text-to-speech audio, saved progress, subscriptions, transactional emails, and support.
• For legitimate interests: security, abuse prevention, debugging, service reliability, minimized first-party usage records, aggregate cookieless website measurement through Plausible, prompt and product improvement, customer support, understanding whether essential features work, and administering referral credit and partner commissions from affiliate links.
• With consent or at your request: microphone/browser permissions, playing an embedded YouTube video, optional marketing emails where required, optional testimonials, and other optional features where consent is the appropriate basis.
• For legal obligations: tax, accounting, payment records, fraud prevention, responding to lawful requests, and enforcing legal claims.

4. AI, speech, and human review

Copycat Cafe uses AI and speech vendors to power language-learning features. When you use chat, pronunciation, transcription, or audio-generation features, we may send the relevant text, chat context, audio, transcript, target phrase, and metadata to the vendors listed below so they can provide that feature.

Authorized Copycat Cafe staff and contractors may access production chats, transcripts, recordings, scores, logs, screenshots, and related metadata when reasonably needed for support, debugging, security, abuse investigation, reliability, quality assurance, prompt improvement, model evaluation, and product improvement. This may include controlled internal engineering tools and coding-agent tools such as Amp. We limit this access to people with a work-related need, require confidentiality, and tell staff to include as little personal data as practical in internal review threads. We treat this as running and improving the service, not as training an AI model.

We do not sell your chats, recordings, or transcripts. We do not use them to train Copycat Cafe’s own AI model. For production AI and speech features, our baseline is to use vendors, routes, and account settings that say customer content is not used to train or improve general AI models without our opt-in. This is not a promise that providers keep no data at all: they may still process or keep limited data for security, abuse prevention, reliability, support, legal compliance, or billing, depending on the provider, route, model, account settings, and contract terms. Do not submit highly sensitive personal data in practice conversations unless it is necessary.

We do not use voice or audio data to identify you as a unique person, authenticate you, or create a biometric identifier. We use audio and speech data to provide transcription, pronunciation feedback, generated audio, support, debugging, quality assurance, and service improvement as described in this policy.

Voice recordings and free-form practice content may reveal sensitive details such as accent, health, ethnicity, religion, political opinions, or other personal information if you choose to say or type them. Copycat Cafe does not need that information for ordinary language practice, so please avoid including highly sensitive personal data in recordings, chats, prompts, or support messages unless it is necessary for your request.

What happens when you submit a recording: we may process the recording in Copycat Cafe and send it to the speech provider needed for the feature. For speech-to-text, we may send audio to Soniox. For pronunciation scoring, we may send audio, the target phrase, language or dialect settings, and related metadata to Langcraft. Our current production pronunciation-scoring flow stores the score, transcript or analysis, target text, feedback metadata, and bounded recording diagnostics linked to the pronunciation attempt so we can understand whether audio conditions affect scoring; it does not attach the raw pronunciation audio recording to your account after the request completes. Text-to-speech providers receive text to be spoken and voice or audio settings, not your own pronunciation recording, unless a feature specifically tells you otherwise.

Public microphone check: the basic check at /microphone-check analyzes and plays a short sample entirely in your browser. The sample and its diagnostic measurements are not uploaded to Copycat Cafe. Closing, refreshing, or resetting the page discards the sample from the checker.

Can you use Copycat Cafe without uploading voice? Yes, you can read and listen to lessons, review content, and type in chat without giving microphone access. Features that need your speech—pronunciation scoring, spoken chat, and speech-to-text—require microphone audio because the app cannot score or transcribe speech it never receives.

Different AI and speech providers handle data differently. We choose production provider routes and account settings intended to meet a simple baseline: no training or model improvement on Copycat Cafe customer content. For OpenRouter routes, this means we send requests with data collection disabled. That is lower than a strict zero-data-retention requirement, but it is intended to avoid routing customer prompts and outputs to endpoints that use them for training or model improvement. Provider retention, abuse monitoring, security review, deletion periods, and processing locations may depend on the provider, route, model, account setting, and contract terms. We do not promise that every provider has identical retention rules or that every route is zero-retention.

We do not authorize AI or speech providers to use Copycat Cafe customer chats, transcripts, or audio recordings to train, fine-tune, or improve general-purpose models unless we explicitly opt in, update this policy, and have an appropriate legal basis, including consent where required.

Automated decisions

Copycat Cafe does not use your learning data to make decisions that produce legal or similarly significant effects about you solely by automated means. AI scores, corrections, and feedback are learning aids. Paddle and other payment/security providers may use automated systems for fraud prevention, payment authorization, tax, or billing decisions under their own terms and privacy policies.

5. Vendors and recipients we actively use

We use processors, subprocessors, and independent controllers to run Copycat Cafe. Based on our production code paths and production data, these are the active categories we currently rely on:

Some vendors act as our processors or subprocessors and process personal data under our instructions. Some vendors, such as Paddle when acting as merchant of record or reseller for payments, may act as independent controllers for their own legal, tax, fraud-prevention, payment-compliance, and support purposes. Where a vendor acts independently, its own privacy notice also applies.

We also use first-party product analytics in our own application database to understand visits, events, feature usage, and learning progress. For Copy Passes, this can include whether an eligible in-app offer was shown or opened, whether the sender used the copy or device share control, and whether an invitation reached the claim and activation steps. We do not put the private Copy Pass link or token, recipient contact details, share destination, or message content in analytics. Historical records may contain metadata from providers we no longer use for new production requests; those providers are not listed here unless we actively send them new data.

AI and speech providers may change by feature, model route, fallback, experiment, or availability. We keep internal records of active production providers and update this policy when a new provider or processing purpose materially changes how personal data is handled. We will not materially expand the use of chats, recordings, or transcripts for model training without an appropriate legal basis and notice, and consent where required.

6. Cookies, local storage, and similar technologies

• Strictly necessary storage: we use an encrypted first-party session cookie for login, security, signup and checkout continuity, application preferences, subscription access, and referral attribution when you arrive through an affiliate link. We also use limited first-party storage for security, interface state, checkout coordination, and installed-app functionality. These are used to provide or secure the service rather than to build an advertising profile.
• Analytics: Plausible records public page paths, normalized private route patterns, engagement time and scroll depth, allowlisted UTM campaign labels, sanitized referring paths, and curated conversion goals without cookies, browser storage, durable identifiers, or cross-site profiling. Outbound-link measurement is disabled. Its anonymous visitor measurement rotates daily. Our first-party analytics records only allowlisted, meaningful product and service events in our own database. Optional anonymous events use a random daily key held in our encrypted session; we do not create a durable visitor profile or join that key to an account. Event records may contain normalized paths and bounded categories or identifiers from a fixed catalog, but not full IP addresses, raw user agents, precise coordinates, arbitrary query parameters, bearer values, chats, recordings, or transcripts. Signed-in product and commercial history may be associated with the account.
• Affiliate attribution: Affonso runs on our pages to detect and preserve affiliate referrals. When you follow an affiliate link, it stores its referral ID in the first-party affonso_referral cookie for up to 60 days and may also carry the ID in the URL and our necessary encrypted session. If you sign up, we retain attribution with your account and pass the needed referral ID to Affonso and Paddle to credit the partner, calculate commissions, prevent fraud, and resolve attribution disputes.
• Operational and billing tools: Sentry provides minimized error and performance diagnostics. Paddle Retain / ProfitWell may run only for signed-in users with Paddle customer records to provide payment recovery or subscription-related notices.
• No advertising trackers: we do not currently use Meta Pixel, Meta Conversions API, retargeting pixels, or cross-site behavioral advertising trackers. The Affonso cookie is limited to first-party affiliate attribution and commission administration rather than cross-site behavioral advertising.
• Embedded YouTube videos: blog video previews are served without contacting YouTube. If you choose “Accept YouTube and play video,” your browser connects to YouTube and YouTube may use cookies, local storage, or similar technologies for playback, security, preferences, measurement, and other purposes described in Google’s Privacy Policy. We do not remember this choice, so another embedded video requires another activation.

7. International transfers

We are based in Cyprus, but our vendors may process data in the European Economic Area, the United Kingdom, the United States, and other locations. Where required, we rely on appropriate safeguards such as data processing agreements, Standard Contractual Clauses, adequacy decisions, or other lawful transfer mechanisms. We do not currently promise EU-only processing for all AI, speech, analytics, or infrastructure vendors.

AI model routing, fallback providers, and speech-processing infrastructure may mean the exact processing location varies by feature, provider, route, and availability. When personal data is transferred from Cyprus or the EEA to a country without an adequacy decision, we rely on safeguards such as Standard Contractual Clauses, data-processing terms, access controls, encryption in transit, and data minimization where required.

8. Retention

We keep personal data only as long as needed for the purposes described above, unless a longer period is required for legal, tax, accounting, security, fraud-prevention, backup, or dispute-resolution reasons. Our current retention practices are:

• Account profile and settings: kept while your account exists, and then deleted or anonymized within a reasonable operational period after account deletion unless we need to keep limited records for legal, security, fraud-prevention, or dispute purposes.
• Copy Pass and promotional-access records: kept for the link and access period and afterward as reasonably needed to enforce eligibility and offer limits, prevent abuse, answer support requests, and resolve disputes. If you delete your account, we detach these records from your user and account. We may retain a keyed, non-reversible digest derived from the normalized account email so the same person cannot repeatedly claim a one-use promotion. The digest is not the email address and is not used to contact you.
• Public result links: new links are available for 30 days unless you revoke or delete them sooner. Revoked and expired records remain in your account controls until you delete them. Deleting removes the result and attached share image from the live service, subject to limited backup, security, fraud-prevention, and legal retention described below.
• Learning progress, chat history, answers, flashcards, scores, pronunciation records, and assessment-linked recording diagnostics: generally kept while your account exists so the product can work, you can review your progress, and we can investigate scoring reliability. You may request deletion as described below.
• Subscription, invoice, tax, and accounting records: kept as required for legal, tax, accounting, fraud-prevention, and dispute purposes.
• Audio and transcripts: raw pronunciation audio is processed to provide scoring and is not attached to your Copycat Cafe account after the current production request completes. We may keep transcripts, target text, scores, analysis, and feedback while your account exists so the product can work and you can review your progress. Provider-side audio is deleted after processing where supported by the integration or retained according to the provider’s terms and our account settings.
• Support messages: kept for a reasonable support-history period unless needed longer for disputes, safety, abuse prevention, or legal obligations.
• Operational logs, analytics events, and error reports: legacy Rails/Ahoy visits are generally kept for 13 months, ordinary legacy marketing/product analytics events for 25 months, and legacy debug analytics events for 90 days. Phoenix optional analytics events are generally kept for 395 days and necessary first-party service events for 790 days. Historical aggregate daily totals and bounded acquisition dimensions may be kept without visitor or user identifiers after the older Rails records expire; debug analytics are not included in those rollups. Canonical account, learning, and commercial event mirrors may be retained with the underlying account, product, billing, legal, tax, fraud-prevention, support, or dispute records. Other operational logs and error reports are kept for limited operational periods unless needed longer for security, debugging, abuse prevention, or reliability.
• Backups: deleted or overwritten on a rolling schedule unless preserved for security, continuity, or legal reasons.
• Affiliate referral identifiers: URL and session continuity lasts through the referral and signup journey. If you sign up, attribution is retained with your account. Paddle/Affonso signup and conversion records may be kept for partner commission, fraud-prevention, accounting, and dispute purposes.

If you ask us to delete data, we will delete or anonymize what we reasonably can, subject to legal, tax, payment, security, fraud-prevention, backup, and dispute-resolution obligations.

Self-service account deletion: you can request deletion from your profile. If an account you own has an open Paddle Billing subscription or trial, the deletion flow requires a separate confirmation before it cancels billing immediately, ends access, and deletes the account. Account deletion does not request or issue a refund, so contact support before deleting if you want to request an available money-back guarantee. If automatic cancellation is unavailable, you must cancel the subscription from Billing first. If an account you own still includes other learners, you must transfer or remove that account first so deleting one person does not delete someone else’s access or records. Once those blockers are cleared, deletion removes the live user profile, learning activity, chat history, public result records and their stored share images, and account attachments; revokes unclaimed Copy Pass links; and detaches or anonymizes limited operational, promotional, and billing history that must remain for accounting, security, fraud prevention, support, or legal claims. Paddle and other independent controllers may retain payment and tax records under their own legal obligations.

Deleting your Copycat Cafe account or requesting deletion may not immediately remove copies already processed by vendors from transient logs, security logs, rolling backups, abuse-monitoring systems, email delivery records, or legal-compliance records. Live application data is removed through the deletion process, while backup and limited operational copies age out on their normal restricted schedules unless they must be preserved for security, continuity, payment, tax, dispute, or legal reasons. Vendor-side deletion and retention are governed by our contracts, provider terms, technical settings, and legal obligations. We minimize what we send to vendors and use shorter-retention or no-training settings where available.

9. Security

We use technical and organizational measures designed to protect personal data, including HTTPS, access controls, authentication, vendor security controls, monitoring, and limited internal access. No online service can guarantee perfect security.

10. Your rights

If GDPR or similar laws apply to you, you may have rights to access, correct, delete, restrict, object to processing, receive a portable copy of your data, withdraw consent, and lodge a complaint with a supervisory authority. In Cyprus, the supervisory authority is the Office of the Commissioner for Personal Data Protection.

You can manage public result links directly from Privacy & sharing in your account settings. To exercise other rights, contact [email protected]. We may need to verify your identity before acting on a request.

You can object to processing based on legitimate interests, including analytics, product-improvement, or security uses, where applicable, and you can object to direct marketing at any time. You can also withdraw consent where consent is the basis. Contact [email protected] to exercise these rights. Some data may still be needed to provide the service, maintain security, comply with law, process payments, prevent fraud, administer earned affiliate commissions, or resolve disputes.

11. Family, school, and administrator visibility

If you join through a family, school, team, or similar plan, the person or organization managing that plan may receive limited account-management information such as your email address, invitation status, seat status, subscription access, and usage needed to administer the plan.

If we offer educator, school, classroom, or team dashboards, the relevant administrator may be able to see learner progress, usage, scores, assignments, or similar information, as described in the product or separate school/team terms. We do not offer those expanded dashboard disclosures unless they are described at or before signup for that plan.

12. Children

You must be at least 16 years old to use Copycat Cafe. We do not knowingly collect personal data from children under 16. If you believe a child under 16 has used the service, contact us so we can review and delete the data where appropriate.

13. Keeping this policy current

Copycat Cafe changes as we add languages, features, vendors, AI models, speech providers, analytics, billing flows, support tools, and evaluation processes. We maintain this policy as an operational document: when we add or materially change a production vendor, model route, cookie/local-storage use, production-data review process, or data-retention practice, we should review this policy and update it before or alongside the product change.

14. Changes

We may update this Privacy Policy when our service, vendors, or legal obligations change. If a change is material, we will take reasonable steps to notify users or request acceptance where appropriate.