# Personal Data Processing Policy

Version of 18 October 2026. Effective 18 October 2026.

## 1. Operator and scope

The Operator is **Pavel Igorevich Shepshelevich**, a professional income tax payer, Russian INN **772990579929**. Contact: [support@open-learn.ai](mailto:support@open-learn.ai).

This Policy covers visitors, users of the website, chat, studios, mini apps and Telegram bot, gallery authors, mailing recipients and people contacting support. Service and payment terms are in the [offer](/offer) and [terms](/terms).

## 2. Data

- Account: email, name and avatar if provided, language and settings; passwords are stored as hashes. Google, Yandex ID and Telegram supply identifiers and available profile fields when those sign-in methods are used. Their passwords are not supplied to OpenLearn.
- Sign-in protection: TOTP secret and backup codes if enabled, public passkey keys and security events. Private passkey keys stay on the device.
- Requests and results: messages, prompts, files, images, video and audio, history, extracted text, memory and personal instructions.
- Usage and payments: model, processing volume, technical operation identifiers, cost, balances, payment amount, currency and status. Payment details are entered with the payment provider; the Service does not store full card numbers.
- Technical information: IP address, browser and device, cookies, error and security logs. Signup attribution may include campaign tags, referral code, referring domain and landing page.
- Support and choices: correspondence and records of consent, publications and their cancellation to the extent described below.

Do not upload special-category data, health information, biometric identification data, protected secrets or other people's personal data without a lawful basis. Requests are processed to perform the selected function; automated processing does not make them anonymous.

Derived data including extracted text, search representations and vectors, temporary files and technical caches may be used for the relevant functions. Service AI stages such as titles, memory and search refinement may also receive necessary context.

## 3. Purposes and bases

| Purpose | Data | Basis |
|---|---|---|
| Account, requests, history, memory and personalization | Profile, content, settings | Necessary contractual processing under Article 6(1)(5) of Russian Law 152-FZ |
| Balances, payments and required transaction documents | Payment and usage records | Contract and applicable legal obligations |
| Security and abuse prevention | Technical data and events | Applicable obligations and legitimate interests subject to individual rights |
| Support and service notices | Contact details, enquiry, operation information | Handling enquiries, contract and legal duties |
| Advertising email | Email, name and language if available | Separate optional [consent](/consent-marketing) |
| Gallery author display | Name and avatar | Separate publication choice; [dissemination terms](/consent-gallery) |
| Optional visit statistics | Browser and visit data | Choice in cookie settings |

Consent does not replace a contractual basis or the Operator's legal duties. The current email signup form separately requires [processing consent](/consent); the advertising option is independent and off by default. Signup attribution is described in section 10 and is separate from permission for Metrica.

## 4. Processing operations

Collection, recording, organization, accumulation, storage, correction, retrieval, use, access, transfer, restriction, deletion and destruction are performed for the relevant purpose. Processing is mainly automated; the Operator handles support. Publication disseminates selected information to an unlimited audience.

## 5. Transfer and safeguards

Account safeguards include sessions, password hashing and available sign-in protection. The public website uses HTTPS. Access to private files is checked; some files are delivered using temporary signed links. Avatars and explicitly published materials are available through their public routes. Do not share signed links with others.

These measures are not a promise of absolute security, certification or compliance with every foreign legal regime. See [security](/security).

## 6. Storage

The Service's primary deployment is intended to be in Russia: the database and server application in Moscow, object storage in St Petersburg through Timeweb Cloud. This does not mean that every processing stage occurs only in Russia: external APIs and technical intermediaries participate in the transfers below. Database backups are encrypted before external copying. Different copies have different retention rules (section 11).

## 7. Recipient categories

- **Timeweb Cloud**: application, database, file and backup hosting.
- **OpenRouter and providers of selected models**: AI requests across modalities, including speech transcription and synthesis. Depending on the function, context history, content and file links, memory, instructions and technical user/request identifiers are transmitted. A model name does not always identify the ultimate processor: routing among providers is possible.
- **Technical network intermediaries**: external API traffic and Telegram interactions. Their involvement does not imply that content is inaccessible at every stage.
- **Resend**: service and, with separate consent, advertising email; address, required template information and delivery records.
- **Payment providers** available in the interface (the code supports Platega, YooKassa, Stripe and Heleket): payment, amount, currency and technical identifiers. Availability depends on the enabled method.
- **Telegram**, **Google**, **Yandex**: bot, Mini App or the selected sign-in method. These services have their own rules; deleting OpenLearn does not automatically delete their copies.
- **SearXNG and connected search engines**, or OpenRouter's search partner: search queries that may include message fragments.
- **Websites whose pages the Service reads**: the URL including parameters, server IP and technical request headers. Page text may be sent to the model. Reference tools including Wikipedia and open-meteo.com receive necessary query data, such as a city name.
- **Yandex Metrica**: visit statistics after permission (section 10).
- **Visitors to shared links and the gallery**: material you explicitly made public. Copies held by visitors and search engines can remain outside the Service's control.

Disclosure to competent authorities may occur where required by law. Recipients receive information for their function; avoid unnecessary personal information in requests and file names.

## 8. International transfers

Model processing, email, third-party interactions, search and transit through foreign infrastructure may transfer information outside Russia. Geography depends on the function, ultimate provider and its contractors; not every route is limited to one country.

A contractual basis does not replace Article 12 of Russian Law 152-FZ: recipient information, required notifications, restrictions and prohibitions must be addressed separately. This Policy and a user's action do not themselves establish that the Operator has fulfilled those duties. Do not send information that must not leave your jurisdiction if foreign processing is unacceptable.

## 9. Model training

OpenLearn does not train its own models on user conversations. Data is sent to external AI services to perform requests. Recipients' use, technical retention and training depend on the terms and settings of each route. A universal prohibition on training or zero retention is not guaranteed across every model, modality and fallback. Do not submit information that requires an unverified guarantee of this kind.

## 10. Cookies and statistics {#cookies}

Cookies and localStorage support sessions, sign-in protection, language, theme, preferences and the analytics choice. External sign-in state is time limited.

**Yandex Metrica** loads only after “Allow analytics” when the counter is enabled in this website build. “Necessary only” or closing the banner does not permit Metrica. It may use `_ym_*` cookies, a browser identifier and information about pages, device and key actions. This is not a promise of anonymity. Session replay and click maps are disabled. Chat content and files are not sent as Service events. The Service does not send hits or goals in the admin panel or Telegram Mini App.

Use **“Cookie settings”** in the footer to change the choice. On refusal, Metrica cookies and storage managed by the Service are cleared and an active page reloads. Previously transmitted information is not automatically recalled. Choice, time and banner version are stored in the browser, not in a server consent archive. See [Yandex's policy](https://yandex.ru/legal/confidential/).

**Signup attribution** is separate: campaign tags, referral code, referring domain and landing page may be remembered on first visit for up to 30 days and sent to a new account after sign-in. Refusing Metrica does not itself disable this mechanism. Optional analytics and referral reward accounting are not the same purpose.

## 11. Retention and deletion

| Data | Storage and deletion |
|---|---|
| History, memory, instructions and generations | Stored for account functions; individual objects can be managed in the interface |
| Temporary chat uploads | The code provides an availability window of up to 30 days from the original upload; extracted text may remain in history separately. Physical deletion depends on storage rules |
| Avatars and durable studio files | Outside the temporary chat window; removed by separate interface actions |
| Account | “Deactivate account” marks the account deleted and ends the current session; this does not establish physical erasure of every record, file or publication |
| Backups | Deleted information can remain until rotation; the supplied configuration is 30 days locally,60 days for external copies plus 14 days for old versions. This is not a complete account erasure deadline |
| Technical caches and bot memory exports | When the relevant functions are used, these can contain text, derived data and metadata; ending processing must account for these copies as well as the main account |
| Payments, enquiries, logs and consent records | Retained for the relevant purpose and mandatory legal periods; ask the Operator for the basis and period applicable to a specific request |

Publications, shared links and individual data can be managed in the interface. Account deletion does not replace unpublishing all materials or guarantee removal of third-party copies. On a processing termination request the Operator determines the applicable bases and deadlines, stops processing and arranges destruction under the law. A contract or mandatory rule may require some records to be retained. See [help](/help/privacy/retention-and-deletion).

## 12. Rights and enquiries

You may request information, access, correction, restriction or destruction where legally applicable, withdraw consent and contact a competent authority or court. Settings provide an export; its contents are not all technical records and files held by the Service.

Write to [support@open-learn.ai](mailto:support@open-learn.ai). Proportionate account ownership verification may be necessary. Responses and fulfilment follow applicable legal deadlines. Foreign-law rights and applicability require separate assessment; unconditional GDPR/CCPA compliance is not claimed.

---

[Previous edition](/legal/archive/baseline-5edc0b9e/en/privacy.md).