Privacy Policy

1. WHO IS RESPONSIBLE FOR YOUR DATA

This policy explains how personal data is handled when you use Unsaid, an AI simulation product for reflection and storytelling. References to we, us, or our refer to the individual operator identified below.

2. DATA AND ITS SOURCES

• Account data: Email, username, display name, authentication information, and account settings, provided by you or your sign-in provider.
• Submitted material: Chat exports, pasted text, screenshots, descriptions, and optional Telegram imports. This can include names, messages, relationships, and information about people who do not have an Unsaid account.
• Stored persona data: Persona configuration and prompts, extracted memory facts in text, vector embeddings, and ongoing AI conversation history. These are generated from your material and product use.
• Permission records: The requester supplies the participant's Google email, which is stored as a hash for matching, not as anonymous data. We save a token hash, the notice snapshot and version, the requesting account, decision timestamps and the accepting Google-verified account. The original link token is kept by the people using the link, not in the consent table.
• Billing data: Transaction identifiers, purchase history, subscription status, and customer identifiers supplied through Polar. We do not store full payment card details.
• Technical and visit data: Usage counters, request and error information, device category, referring URL, and a daily IP-derived hash for visit measurement. A hash is pseudonymous information, not a guarantee of anonymity.
• Communications: Information you send when you contact us, including privacy requests.

Authentication cookies support sign-in. Browser storage also remembers your theme, Telegram import session, permission-request draft and link while you use the permission flow, and whether a visit has already been counted in the current browser session. Embedded media, such as the YouTube walkthrough on About, can involve the media provider when loaded or used. The permission page is excluded from our visit measurement.

3. PURPOSES AND LEGAL BASES

If a particular use relies on consent (Article 6(1)(a)), consent must be specific and can be withdrawn. Accepting the Terms or confirming permission to upload material is not a blanket consent to every use of personal data.

Account information and relevant conversation context are needed to provide the features you request. You can choose not to submit material, but analysis and persona features need enough context to work.

4. DATA ABOUT OTHER PEOPLE

Information about someone who has not registered usually comes from material another user supplies, such as a private chat export or screenshot. Their data can appear in submitted text, persona configuration, memory facts, and conversation context.

Your service contract with us is not a legal basis for processing another person's data. Only submit identifiable information where that person's valid consent covers this use and disclosure to the relevant service providers, or another applicable lawful basis has been established. Remove identifying details when you do not have a basis to share them.

The current creation flow requires a permission link for one living adult participant and one private AI persona. The requester shares the link personally. The participant can read the notice without signing in and can accept using the Google-verified email specified by the requester, or decline. Opening a link or remaining silent is not consent. The requester cannot accept their own request. Selected sources stay in the requester's browser until confirmation; analysis, screenshot transcription and Telegram import require active permission on the server.

This permission covers ordinary information from exchanges between the requester and that participant. It does not cover other people's data, sensitive information or minors, which the requester must exclude. Google confirms control of an email account; it does not prove who authored every message. We do not automatically establish that every person mentioned in a file has consented. Older personas without a recorded participant permission are blocked from further conversation and memory processing.

Where Article 14 GDPR applies, we are responsible for giving the affected person the required information, including the source of their data. Notice is normally due within one month, or earlier at the first relevant communication or disclosure. An exception needs to be applicable and documented; a user's upload confirmation or this public policy does not automatically replace an individual notice.

Do not submit sensitive information about health, ethnicity, beliefs, sex life, or other special categories unless the applicable Article 9 conditions are met. An ordinary service contract or a general permission checkbox is insufficient for that purpose.

You can contact support@unsaid.ink about your data even if you have never used Unsaid. We may ask for enough information to locate the material and verify your request without exposing another person's private data.

Keep your permission link to withdraw without signing in, or contact us. Withdrawal blocks further use and attempts to remove the linked active persona, conversations, memory and Telegram imports. If removal fails, the receipt reports that it is incomplete and can be retried. A request already in progress with a provider cannot be recalled; the application rechecks permission before saving or returning its output.

5. UPLOAD PROCESSING AND RETENTION

Original files and request processing

The standard file workflow reads selected chat files in your browser and sends their text for analysis. Screenshots are sent as image data for transcription. These workflows do not create a separate original-file archive in the application database.

This describes the application workflow, not guaranteed immediate erasure from every system. Request handling, operational logs, hosting infrastructure, backups, and external AI services can have separate retention. We do not promise that all copies of submitted material are permanently deleted as soon as analysis ends.

Data deliberately saved by Unsaid

• Personas and conversations: Configuration, text memory facts, embeddings, and messages are kept with your account until the relevant persona or account is deleted, or a valid deletion request is completed.
• Telegram imports: Forwarded messages and Telegram identifiers are saved in an import session in the database. Session expiry limits access to the import; it does not by itself erase the database record. Contact us to request removal of import-session data.
• Free and paid access: Canceling a subscription or letting a Day Pass expire changes access limits. It does not itself delete your saved data.
• Permission invitations and receipts: Pending links expire after seven days. Minimal request records, the notice shown and decisions are kept where necessary to demonstrate consent, handle privacy requests or establish, exercise or defend legal claims; expiry does not itself erase the record. Withdrawal does not automatically erase evidence of the earlier choice.
• Operational, visit, and billing records: Retention depends on the purpose, applicable legal requirements, and provider settings. There is no single verified deletion deadline for all such records.

Deletion controls affect records in the application. They do not themselves erase copies already processed by third parties. Backup and provider retention can differ from removal in the active application database.

6. SERVICE PROVIDERS AND AI DATA PRACTICES

• OpenRouter and routed model providers: Receive relevant submitted material and conversation context for transcription, analysis, memory extraction, embeddings, and generated replies. A selected model and routing can change the actual recipient.
• Supabase: Stores account records, persona configuration, text memory facts, embeddings, saved conversations, Telegram import sessions, and visit records.
• Polar: Processes paid purchases and subscriptions, with its own obligations for payment and customer information.
• Hosting and integrations: Infrastructure providers handle requests and diagnostics. Telegram and embedded media providers also process information when you use those features.

OpenRouter and individual model providers have separate retention and data-use rules. Some providers may retain submitted content or use it for model training or improvement, depending on the provider, settings, and agreements. This policy makes no universal promise of zero data retention or exclusion from training.

We use OpenRouter's Zero Data Retention controls for Anthropic model requests and for the separate model used to generate memory embeddings. Embedding requests explicitly require ZDR routing. OpenRouter restricts these requests to endpoints eligible under its ZDR policy; an embedding request fails if no eligible endpoint is available. ZDR concerns retention at the routed model provider. It does not remove the generated embeddings, text memory facts, persona configuration, or conversations deliberately saved by Unsaid. OpenRouter's own optional input and output logging, hosting logs, and backups are separate controls.

See OpenRouter's privacy policy and its provider routing and data controls. Contact support@unsaid.ink for information about the model and providers used for your request.

7. INTERNATIONAL PROCESSING

Our service providers can process personal data outside Poland and the European Economic Area, including in the United States. Locations depend on the hosting configuration, model routing, and providers involved.

Transfers outside the EEA require an applicable legal mechanism. Adequacy decisions and Standard Contractual Clauses are possible mechanisms; using a provider does not itself establish that a particular safeguard applies. This policy does not assert that every transfer is covered by executed SCCs.

For information about actual processing locations and available transfer safeguards, contact support@unsaid.ink. The relevant recipient, agreements, and configuration need to be verified for the specific transfer.

8. YOUR DATA PROTECTION RIGHTS

Where applicable under GDPR, you can request:

• Access to your personal data and information about its processing.
• Rectification of inaccurate or incomplete personal data.
• Erasure of personal data where the legal conditions are met.
• Restriction of processing in the circumstances provided by law.
• Portability of eligible data in a machine-readable format, and transfer to another controller where technically feasible.
• Objection to processing based on legitimate interests, subject to the applicable conditions.
• Withdrawal of consent for a use that relies on consent, without affecting the lawfulness of earlier processing.
• A complaint to a supervisory authority, including where you live, work, or believe an infringement occurred. Poland's authority is the President of the Personal Data Protection Office (UODO).

Send a request to support@unsaid.ink. An Unsaid account is not required. We may request proportionate identity verification. GDPR requests should be answered without undue delay and normally within one month; if a lawful extension is needed, we will explain it within that first month.

9. AI ANALYSIS AND AGE LIMIT

AI analysis creates persona characteristics and retrieves memories to generate simulated replies. It can infer information incorrectly. We do not use persona responses to make decisions producing legal or similarly significant effects about you under Article 22 GDPR.

Unsaid is intended for adults aged 18 and over. Do not submit material involving minors. If you believe a child's personal data has been provided, contact support@unsaid.ink so we can investigate and handle an appropriate removal request.

10. POLICY UPDATES

Updates will be posted here with a revised date. Material changes to data use may require additional notice or consent under applicable law. Publishing an update does not by itself supply a new legal basis for processing.