← All legal documents
This document has not yet been translated into RU. The English version is shown below and remains authoritative.

Version 1.0.0 • effective Fri May 15 2026 00:00:00 GMT+0000 (Coordinated Universal Time)

BRAIN Addendum

This addendum applies when you use BRAIN — NAL's AI agent for working with your personal documents, calendar, projects, and commands. BRAIN is accessible:

  • as a module embedded in cloud.nal.digital (the OpenCloud-based file storage interface),
  • at the stand-alone domain brain.nal.digital (during transition),
  • through the brain-guild mobile app (Android, hybrid RAG + knowledge graph + social),
  • via the Telegram bot.

It is part of the Terms of Service and incorporates the AUP, Privacy Policy, and Content License.

Effective date: 2026-05-15

1. What BRAIN Is

BRAIN is an AI agent that, with your permission:

  • reads, indexes, and answers questions about files you store in cloud.nal.digital (OpenCloud);
  • creates, edits, and organises projects, schedules, and reminders;
  • uses MCP servers to act across other NAL Services (Guild, Studio, Music) when you ask it to;
  • sends messages through the Telegram bot integration;
  • maintains a knowledge graph over your data to support retrieval- augmented generation (RAG).

2. Authority You Grant BRAIN

2.1. When you give BRAIN access to a file, project, calendar, or external account (Telegram, OpenCloud), you authorise BRAIN to read and act on that data within the scope of your instruction.

2.2. You are the principal of every action BRAIN takes. If BRAIN sends a message, schedules an event, transfers a file, posts to Guild, or spends credits from your wallet, it is your action under these Terms.

2.3. Pre-authorisation. Certain actions require you to confirm explicitly each time, including:

  • spending above per-session or per-day caps you set;
  • posting content to a public guild;
  • sending messages to people you have not previously interacted with;
  • deleting files or projects.

2.4. You may revoke any authority you have granted BRAIN at any time from your account settings. Revocation takes effect immediately for new actions; in-flight actions complete or fail according to their state at the moment of revocation.

3. Files, OpenCloud, and Indexing

3.1. BRAIN integrates with OpenCloud at cloud.nal.digital. Files you upload to OpenCloud are stored in NAL infrastructure (Hetzner Nuremberg, MinIO / S3) under your account.

3.2. Indexing. Files you place in OpenCloud (or sub-folders you explicitly enable) are indexed by:

  • extended-rag for knowledge-graph extraction (entities, relations),
  • embeddings-builder for vector search,
  • graph-service (Neo4j) for graph storage.

The indexes are scoped to your account. NAL does not pool indexes across users.

3.3. Auto-indexing scope. By default, BRAIN auto-indexes only the folder you have designated as the BRAIN workspace. Other OpenCloud folders are not indexed unless you opt them in.

3.4. Deletion. When you delete a file from OpenCloud, the corresponding embeddings, graph nodes, and cached LLM contexts are deleted from the indexes within 24 hours. Backups roll on a 30-day cycle (see Privacy Policy).

4. AI Inference for BRAIN — Third-Party Visibility

⚠️ Your prompts go to third-party AI providers. When BRAIN answers you, the relevant slice of your data — the file excerpts, calendar entries, or message context BRAIN selects — is sent to the third-party AI provider you selected (OpenAI, Anthropic, Google, etc.). NAL configures the no-training / zero-retention option where available, but the provider's own privacy policy governs what they see. NAL is not liable for the provider's handling. See /en/legal/partners for the full provider list.

4.1. BRAIN routes LLM inference through the multi-agent proxy to upstream providers (OpenAI, Anthropic, Google, etc.). The prompt sent to the upstream provider contains only the file excerpt or context window relevant to your request — not your full BRAIN workspace, account profile, or unrelated data.

4.2. Where the upstream provider offers a no-training / zero-retention data-handling option, BRAIN uses it by default. Where the provider later changes their policy, BRAIN re-evaluates and may switch providers — your data is not retroactively recallable from the prior provider.

4.3. Voice and dictation. Voice input you give to BRAIN is transcribed by NAL's STT (faster-whisper) running on NAL infrastructure. Voice is not sent to a third-party transcription service unless you explicitly select one.

4.4. Media generation via NAL Net. When you ask BRAIN to generate an image, video, or audio, the job is dispatched to the relevant generation Service (Studio / Music) and may route through NAL Net community workers — third-party operators who see the prompt and output in plaintext. See Studio Addendum §4 and NAL Net Addendum §2. Pick the "hosted only" option in the media tools' model picker for sensitive prompts.

5. Knowledge-Graph Privacy

5.1. The personal knowledge graph BRAIN builds over your files is private to your account. It is not shared with other users, with NAL staff, or with research projects.

5.2. The personal knowledge graph is not used to train AI models.

5.3. Public reach. If you instruct BRAIN to publish material from your knowledge graph (post to Guild, share a document publicly), that specific material then leaves the private scope and is governed by the target Service's terms (e.g., the Guild Addendum).

6. Telegram Bot

6.1. Linking. Linking your Telegram account to BRAIN authorises BRAIN to read and send Telegram messages within the conversation you explicitly bind (one-to-one DM with the BRAIN bot, or specific groups you add it to).

6.2. Limits. BRAIN can only see Telegram messages where the BRAIN bot is a participant. Telegram message contents are processed in real time and are not stored beyond the rolling-context window needed for the conversation, unless you explicitly archive the chat.

6.3. Revocation. Removing the BRAIN bot from a Telegram conversation, or unlinking your Telegram account in NAL settings, revokes BRAIN's access immediately.

7. Calendars, Schedules, and External Accounts

7.1. BRAIN can integrate with external calendars (Google Calendar, CalDAV servers, etc.) when you authorise an OAuth grant.

7.2. Scope of grant. BRAIN uses the minimum scope needed for the feature you enabled (read for "show me my schedule"; read+write for "create an event"). You can revoke the grant in the external provider's settings or in BRAIN's account settings.

7.3. External data. Data fetched from an external account is held only as long as needed to answer the current request, plus the rolling context window. We do not retain external calendar data for indexing beyond active sessions.

8. Pricing and Wallet

8.1. BRAIN actions priced per call (LLM inference, media generation, external Service calls) are charged to your wallet at the prices published at /en/legal/pricing.

8.2. Estimates. Before a large action, BRAIN may display a cost estimate. The final charge can vary within a published margin (roughly ±20%) due to token-count variability.

8.3. Limits you can set.

  • Per-session cap — maximum spend per chat session;
  • Per-day cap — maximum spend across all BRAIN activity in a rolling 24-hour window;
  • Confirmation threshold — actions costing above this require explicit confirmation.

These limits live in your BRAIN settings and apply across all entry points (web, mobile, Telegram).

9. Mobile App (brain-guild)

9.1. The brain-guild mobile app for Android (Kotlin Multiplatform, delivered as a hybrid RAG + graph + social app) is governed by these Terms and the app store's distribution terms (Google Play, or direct APK).

9.2. Device security. The mobile app implements:

  • secure-screen detection (recording / screenshot blocking) for sensitive views;
  • root detection (RootBeer);
  • accessibility-service abuse guard;
  • certificate pinning (where rolled out);
  • encrypted local storage (EncryptedSharedPrefs + AndroidKeyStore).

These are best-effort defences. Rooted devices, modified system images, and physical access can defeat them. You are responsible for the security of your device.

9.3. Offline use. Some content is cached for offline use. Cached content is encrypted at rest and is purged when you log out.

10. AI-Generated Content

10.1. Output BRAIN generates on your behalf belongs to you under Section 5 of the Content License, subject to upstream model provider terms.

10.2. Verification. BRAIN can produce confident-sounding but incorrect statements (hallucinations). For consequential actions (legal, medical, financial, safety-critical) verify before acting.

10.3. Citations. Where BRAIN cites a source from your knowledge graph, the citation links back to the source file or graph node. Open the citation to confirm.