AI policy — Diarynchy

This is an unofficial translation for informational purposes. The Russian version is legally binding.

1. About this document

This AI Policy describes how the Diarynchy service uses artificial intelligence to analyze diary entries, what data models receive, and what limitations apply to users.

2. Which models are used

Analysis of user entries is performed using language models hosted on Yandex.Cloud LLC infrastructure (Yandex AI Studio service) and on the Operator's infrastructure in the Russian Federation.

Language models from the YandexGPT family via Yandex AI Studio (Yandex.Cloud LLC), as well as machine learning models hosted on the Operator's infrastructure in the Russian Federation, are used.

Personal data of users is not transferred to foreign organizations.

All user data is initially stored and processed in the Russian Federation.

3. What data models receive

Analysis uses the minimum necessary amount of data; only the fragment of an entry required for the requested feature is transferred.

Models receive only the text fragment necessary for the requested feature. Account identifiers, e-mail addresses, and other direct personal identifiers are not sent to models.

Full accounts, entire entry history, and payment data are not sent to models.

4. Automated Guardrail review

Before transferring an entry fragment to AI systems, the Service may perform automated Guardrail review. Guardrail is used to mask direct identifiers, limit transfer of excessive information, detect clearly dangerous data, and verify required consents.

5. Special categories of personal data and AI

If an entry contains information relating to special categories of personal data, AI analysis of such an entry is performed only with two separate consents: consent to processing special categories of personal data and consent to automated processing using AI.

6. Retention of analysis results

Automated analysis results are stored with the corresponding diary entry in the user's account in the Russian Federation.

Detailed retention periods are in the Data Retention Policy.

7. Use of data for model training

The Service does not use users' personal data to train its own artificial intelligence models.

When processed through Yandex.Cloud LLC services, data is handled under Yandex Cloud terms; the Operator does not use users' personal data to fine-tune its own models.

Yandex Cloud processing terms: https://yandex.ru/legal/cloud_dpa

8. Who has access

Entry processing for AI features is performed automatically. Operator staff do not review entries in bulk.

Access to individual texts is possible only in limited cases: fixing failures, investigating security incidents, or handling a support request — only to the extent necessary.

When Yandex.Cloud LLC services are used, personal data processing is entrusted under the Yandex Cloud terms and Data Processing Addendum (DPA) accepted by the Operator when registering in Yandex Cloud. Details: https://yandex.ru/legal/cloud_dpa

9. AI limitations

AI results are auxiliary and may contain errors, inaccuracies, incomplete or outdated information.

Results are not a medical conclusion, psychological counseling, psychotherapy, legal advice, or financial recommendation.

Analysis is performed automatically without making legally significant decisions regarding the user.

The User independently makes decisions based on the information received.

10. Disabling AI features

The User may use the Service solely as an electronic diary without intelligent analysis features. AI features can be disabled in app settings.

When AI features are disabled, entries continue to be stored; automated analysis of new and existing entries is not performed.

Withdrawal of consent to automated processing using AI can be sent to k.kulakov@diarynchy.com.

11. Related documents

Also see: