Omnia One AI Privacy Policy

Effective date: 2 November 2025

This Privacy Policy explains what personal data the Omnia One AI application processes, for what purposes and how it is protected. The controller is Cristian Bucioaca, contact: cristianbucioaca@omniaoneai.com.

1. Roles under GDPR

• The Provider acts as the data controller for personal data submitted by users or generated while using the application.
• Third-party platforms (Supabase, Google Cloud, OpenAI, Anthropic, ElevenLabs, Stripe) act as processors or independent controllers depending on the service they deliver.
• Data are shared with third parties only to the extent necessary to provide the Service.

2. Categories of Processed Data

1. Identification and login data: e-mail, hashed password, internal user ID.
2. Profile information: optional display name and nickname, preferred localisation.
3. Chat content and attachments: text messages, system metadata, files uploaded by the user (PDF, images, documents), AI-generated outputs.
4. Technical metadata: device ID, timestamps, browser version, error logs.
5. Usage metrics: message counts, token usage, subscription activity (if enabled).
6. Payment data: handled by Stripe; the application stores only reference identifiers (customer_id, subscription_id).
7. Voice data: short-lived audio streams for speech-to-text or text-to-speech conversions; they are not stored after processing.

3. Purposes and Legal Basis

4. Data Sharing

• Supabase (EU regions): authentication, database (profiles, chats, messages), Storage (attachments, generated-images, generated-pdfs-temp), realtime.
• Google Cloud (EU and US regions per configuration): Vertex AI Gemini, Document AI, Text-to-Speech, Speech-to-Text, Cloud Storage for temporary processing. Credentials are provided as JSON via server-side environment variables.
• Anthropic: Claude API for generative responses and web search. Only the relevant conversation context and supported attachments are transmitted.
• OpenAI: GPT-5 for additional generative capabilities. System instructions and portions of the conversation are sent as required.
• ElevenLabs: Text-to-Speech streaming; only the current text response is transmitted.
• Stripe: subscription management; only tokenised identifiers are stored in the application.
• Google Search / external sources: when web search is enabled, the query is sent without personal identifiers.

No data are sold to third parties or shared for third-party marketing.

4.5. Gmail Integration

When you connect Gmail to Omnia, you can ask AI models to help you read, search, compose, and send emails directly from your conversation. Here's how it works:

What Omnia Can Access:

• Read emails (when you ask the AI to read them)
• Search your mailbox (when you request specific emails)
• Compose and send emails (only when you explicitly instruct the AI)
• View email metadata (sender, subject, date, labels)

How Your Gmail Data Is Used:

• On-Demand Only: Omnia accesses your Gmail only when you explicitly ask (e.g., "show my unread emails", "reply to John's message")
• Direct to AI: Email content flows directly from Gmail to AI models for processing your request
• No Storage: We do NOT store copies of your emails. Email content appears only in your chat and is deleted when you delete the conversation
• No Background Access: We do NOT scan, monitor, or access your Gmail automatically - only when you ask

What We Store:

Only your access token - an encrypted credential that lets you connect Gmail. We do NOT store email content, attachments, or metadata.

Your Control:

• Disconnect Gmail anytime from Google Account settings or Omnia settings
• Delete any conversation to permanently remove all references to your emails
• Review permissions at Google Account Permissions

Note: Omnia's use of Gmail API complies with Google API Services User Data Policy.

5. Storage and Retention

• Supabase: data are stored until the User deletes the chat or the entire account. The interface explains the effects and requires the User to type DELETE and confirm the action directly in-app; no e-mail confirmation is issued. After confirmation the backend delete-account.js callssupabase.auth.admin.deleteUser and cascades profiles, chats, messages and related Storage objects.
• IndexedDB in the browser: the local OmniaChatDB database holds chats for offline access. Users can delete the data in settings or during sign-out.
• Temporary storage: uploaded files processed through Document AI may be stored briefly in Google Cloud Storage; after completion both temporary files and metadata are removed.
• Server logs: retained for up to 30 days for incident response and then anonymised or deleted.
• Invoices and payment records: kept by Stripe in accordance with legal requirements.

6. Security Measures

• All tables in Supabase have Row Level Security rules enforcing auth.uid().
• Storage objects rely on the built-in owner field, allowing only the owner to delete files.
• Access tokens are verified on the backend using the Supabase service role key.
• Passwords are hashed by Supabase according to current security standards.
• All API communication is encrypted (HTTPS).
• When a chat is deleted the application removes Storage files first (deleteStorageFiles), then deletes conversations and sync records.
• Access to critical credentials (Google, OpenAI, Anthropic, ElevenLabs) is limited to server-side code.

7. User Rights

Users have the following rights under GDPR:

1. Access: request a copy of their data (available via account export).
2. Rectification: update profile information (handled by ProfileService.saveProfile).
3. Erasure: delete chats, attachments or the entire account using the provided controls.
4. Restriction: disable synchronisation or cancel subscriptions.
5. Portability: request export of conversations in JSON/CSV format.
6. Objection: object to processing based on legitimate interests.
7. Complaint: lodge a complaint with the Office for Personal Data Protection (www.uoou.cz).

Requests can be sent to cristianbucioaca@omniaoneai.com. The Provider will respond within 30 days.

8. Automated Decision-Making

AI responses are generated using third-party models based on system instructions and conversation history. The Service does not apply automated decision-making that produces legal or similarly significant effects without human oversight. Subscription and billing follow transparent rules.

9. Changes to this Policy

This document may be updated when technical or legal changes occur. Users will be informed about significant changes by e-mail or in-app notice at least 14 days before they take effect. Previous versions are archived and can be provided on request.

This policy reflects the implementation of Omnia One AI as of 2 November 2025.