This is a non-binding English translation provided for your convenience. The legally binding version of this document is the German original („Deutsche Fassung“).

Privacy Policy

1. Data controller

AGAPE INNOVATIONS GROUP INC.
18 King Street East, Unit 1400
Toronto, ON M5C 1C4
Canada
Email: kontakt@hazonlive.com

Corporation incorporated under the laws of the Province of Ontario, Canada. AGAPE INNOVATIONS GROUP INC. is the service provider and the controller within the meaning of Article 4(7) GDPR for the processing of personal data on hazonlive.com.

2. Collection and processing of personal data

We collect personal data when you register, use our services or get in touch with us. This includes:

Name and email address (on registration)
Usage data (chat histories, uploaded documents)
Payment data (processed via the payment service provider Stripe, not stored by us)
Technical data (IP address, browser, device)
Data of connected channels, for Facebook and Instagram e.g. Page ID, Page name, encrypted access token and the ID and username of the Instagram account, as well as IDs and links of published posts and error messages from Meta (details in section 4c)

3. Legal bases

Processing is carried out on the basis of:

Article 6(1)(b) GDPR (performance of a contract)
Article 6(1)(a) GDPR (consent)
Article 6(1)(f) GDPR (legitimate interests)

4. Use of artificial intelligence

Hazonlive uses AI models for text generation, document analysis and image generation. The AI providers are listed in detail as processors in section 6. The following applies:

Your inputs (prompts, conversation histories, image descriptions; for Reels, still frames and the audio track of your video) are transmitted to the AI providers
The AI providers (Anthropic, OpenAI and Google as a backup image model) process the data in the USA; the transfer is based on Standard Contractual Clauses pursuant to Article 46(2)(c) GDPR
Anthropic and OpenAI retain inputs and outputs for up to 30 days for abuse detection, Google for 55 days — longer where content is flagged for safety reasons
In accordance with the providers’ API terms, inputs are not used to train the AI models
All AI-generated content is labelled as such within Hazonlive; published social media posts carry no automatic AI notice (see section 4c)
Further information: AI transparency notice

4a. AI-assisted processing of connected email mailboxes

When you connect an email account (e.g. Gmail) to Hazonlive, the platform regularly retrieves incoming messages and processes them with AI: detecting whether a reply is needed, sorting them into categories (e.g. customer enquiry, appointment, invoice, newsletter) and preparing draft replies. The following applies:

The sender's name and address, the subject line and the shortened message text (at most 4,200 characters, including quoted earlier messages) are transmitted to the AI providers; attachments are neither retrieved nor transmitted (data minimisation), except that with “Read invoices in detail” switched on, Hazonlive retrieves the PDF and XML attachments of invoices and only text taken from them, never a file, is sent to the AI providers, for PDFs together with their file names (see invoice overview)
To learn your writing style, selected messages you have sent yourself may be read; a shortened excerpt is passed to the AI providers as a style example within the prompt (not to train the models). We delete these examples as soon as you disconnect the last connected mailbox of that provider (e.g. Gmail).
So that no enquiry is missed that you have already opened in your mailbox, Hazonlive also retrieves opened messages from the last 48 hours, at most 30 per day. Of these, only enquiries, leads and appointment requests are kept; for all others we keep only a content-free identifier so that they are not processed again. Opened messages are neither marked nor moved in your mailbox.
Invoice overview: from messages recognised as an invoice, receipt or payment reminder, the AI extracts the invoice data (counterparty with name and email address, invoice number, amount, currency, invoice date and due date, payment method, direct-debit date), so that you can see open payments and due dates in Hazonlive. For this, the sender, the recipients, the subject line and the shortened message text (at most 4,200 characters) are sent to the AI providers, together with your company name and the addresses of the mailboxes connected in the workspace so that the AI can tell incoming from outgoing invoices. Attachments are only read if owners or admins switch on “Read invoices in detail” in the settings. Attachments in detail:
- PDF: Hazonlive reads at most 3 PDF or XML attachments of up to 10 MB each from messages recognised as invoices and extracts text from the PDFs, without image recognition. This PDF text is additionally sent to the AI providers, together with the PDFs’ file names, shortened to at most 20,000 characters, solely for this extraction; the document itself is not stored.
- E-invoice: For e-invoices (XRechnung, ZUGFeRD), Hazonlive reads the invoice data from the XML file, without AI; the XML file itself is never sent to the AI providers. The AI additionally receives, besides the message text, a short text built from the header data of the e-invoice (number, dates, amount, payment terms, parties) only if the subject line looks like a payment reminder, a payment confirmation, a receipt or a credit note (e.g. “paid”, “overdue”, “receipt”) rather than a plain invoice, so that the reminder level and payment status are not lost.
- File name: With the switch on, Hazonlive also recognises invoices by the file name of an attachment, without opening it. Only with the switch on, Hazonlive records for new messages whether an attachment name looks like an invoice (yes/no); the name is not stored for this (with Outlook, the names are requested from Microsoft only then, too).
- File reference: With the switch on, for each newly recognised invoice Hazonlive additionally stores the file name, file type, size, format, a checksum of the main invoice file (the file treated as the invoice) and the identifier of the message, never the file. This way the same file can be found again later.
- Sent folder: In that case Hazonlive also reads sent messages whose subject or PDF file name indicates an invoice (going back at most 90 days when switched on) to add your outgoing invoices. For this selection Hazonlive reads only the message headers (such as subject and recipients) and attachment file names, never the text or attachment contents; only a message selected this way is read with its text and attachments as described above. Messages recognisable as forwarded from their subject line (such as “Fwd:”) are not included, replies only if an invoice PDF is attached. Other sent messages are not opened. Sent messages stay in your Sent folder; in Gmail they receive the label “Ausgang”.
- What never happens: Other attachments are never opened. Attachments whose file name contains a payroll, salary or personnel word (e.g. payroll, salary or remuneration statements) are never opened, even if they arrive together with an invoice.
Only the recognised invoice data, the file reference described above, status data (open, paid or cancelled, reminder level), the identifiers, time of receipt and location of the related messages and their conversation in your mailbox and a history (e.g. recognised, filed, marked as paid) are stored, not the message text. In Gmail, the message receives the label “Eingang” (incoming) or “Ausgang” (outgoing) below the invoice label. Only if “Tidy up automatically” is switched on for invoices does Hazonlive move invoices out of the inbox into these folders. Open payment reminders, overdue incoming invoices (except direct debit), invoices Anna is asking you about and invoices whose direction is unclear then stay in the inbox. Nothing is ever deleted, and “Back to inbox” returns a message at any time. Hazonlive does not add messages that were already opened when they were retrieved to the overview on its own. We delete the entries 24 months after they are closed (paid or cancelled), open ones at the latest 36 months after they were recognised, and immediately when you disconnect the last connected mailbox of that provider; if another mailbox of the same provider stays connected, the periods above apply. Payment terms learned per counterparty (sender domain or address and number of days) are deleted 24 months after their last change, even if you disconnect the mailbox earlier. When a payment is coming up or an invoice is overdue, Hazonlive reminds you in the app and also sends owners and admins an email (delivered via Resend) or a push notification, at most once a day and switchable off in the settings. These reminders only contain counts, never the counterparty, amounts or invoice numbers; the daily email summary likewise only shows open payments as a count. At your request, Hazonlive creates a draft payment reminder to your customer in your mailbox and in Hazonlive from the recognised invoice details (fixed template, no AI); it is never sent automatically.
To detect whether you have already answered an enquiry directly in your mailbox, Hazonlive reads only metadata of your sent messages (thread identifier, sending time, recipients, reference headers), never the text (sent invoices with “Read invoices in detail” switched on: see the invoice overview). Only the identifier and time of the detected reply are stored with the enquiry concerned.
Nothing is ever sent, deleted or permanently altered automatically — every reply requires your explicit approval (human-in-the-loop), and automatic sorting can be undone at any time
The legal basis vis-à-vis the senders of your emails is Article 6(1)(f) GDPR (legitimate interest in the efficient handling of incoming business communication); every AI categorisation is stored with a justification and is available for you to view
Access tokens for your mailbox are stored in encrypted form (AES-256-GCM); content from Google services is used exclusively for the functions described here that are visible to you and is not analysed for advertising purposes (Google API Services User Data Policy, Limited Use)
Incoming messages are automatically deleted from Hazonlive after 90 days (answered ones after 24 months; the periods above apply to the invoice overview); your mailbox itself remains unaffected

4b. Use of Google user data (Gmail and Google Calendar)

When you connect a Google account to Hazonlive, we access only the following Google user data and use it solely for the functions described below that are visible to you:

Incoming Gmail messages (read): retrieved so that your AI assistant can detect whether a reply is needed, sort messages into categories, prepare suitable draft replies, recognise appointments mentioned in them and extract the invoice data from invoices for the invoice overview (section 4a). Besides unread messages, this also applies, within narrow limits, to messages opened in the last 48 hours (at most 30 per day, not from the Gmail categories Promotions, Social, Updates and Forums); of these, only enquiries, leads and appointment requests are stored permanently, and Hazonlive does not add them to the invoice overview on its own. The sender, the subject line and a shortened message text are transmitted to the AI providers (for the invoice overview also the recipients, your company name and the addresses of the connected mailboxes), no attachments (data minimisation; the only exception: text taken from attachments of invoices if “Read invoices in detail” is switched on, see 4a).
Sent Gmail messages (read): thread identifier, sending time, recipients and reference headers, never the text (exception with “Read invoices in detail”, see below). This lets Hazonlive detect whether you have answered an enquiry directly in Gmail, so that it is no longer shown as open. Only with “Read invoices in detail” switched on: sent messages whose subject or PDF file name indicates an invoice, together with their PDF attachments, to add outgoing invoices to the invoice overview (section 4a).
Gmail drafts (create): draft replies are stored as drafts in your mailbox. Nothing is ever sent automatically — sending always requires your explicit approval.
Gmail labels and message filing (manage): Hazonlive creates its own label namespace, applies labels to messages and archives them to organise your inbox. Messages are not permanently deleted without your action.
Google Calendar events (read and create): Hazonlive adds appointments to your Google Calendar as a tentative request — when you ask for it in the chat, the reception view or the inbox, when a caller names a specific appointment at the telephone reception, and when an incoming email proposes or confirms a specific appointment with you and the time is free. Automatically adding appointments from emails can be switched off with the calendar switch of the assistant Julia. To check for conflicts, the busy times of your events are read; their titles and contents are neither stored nor evaluated. Hazonlive confirms, moves or removes existing entries only at your request and only if Hazonlive created them itself.
Email address of your Google account: used solely to identify the connected account.

Storage, disclosure and retention: access and refresh tokens are stored in encrypted form (AES-256-GCM, see section 5a); processing takes place on servers in the European Union (Frankfurt). Google user data is disclosed only to the processors listed in section 6, to the extent necessary to provide the functions described, and is never used for advertising or to train AI models. This complies with the Google API Services User Data Policy, including the Limited Use requirements. Incoming messages are automatically deleted from Hazonlive after 90 days (answered ones after 24 months; the periods in section 4a apply to the invoice overview); your Google account itself remains unaffected. You can revoke the connection at any time in the settings.

4c. Facebook and Instagram data (Meta)

When you connect a Facebook Page or an Instagram Business account to Hazonlive under “Integrations” in the settings, we receive the following data via Meta’s programming interface (Graph API). We use it to establish the connection, to publish posts on your Page or Instagram account and to show you the status of your publications. The legal basis is Article 6(1)(b) GDPR (performance of a contract); for error diagnostics, Article 6(1)(f) GDPR (legitimate interest in stable, error-free operation).

Sign-in (only while connecting): we exchange Meta’s sign-in token for a long-lived user token. If Meta does not return a list of Pages, we use the “debug_token” interface to read the details of this token, including the permissions granted and the Pages granted. We use the user token and these details only while connecting and do not store them.
Pages and linked Instagram accounts (read): for every Page you grant in Meta’s login dialog, we retrieve the Page ID, the Page name and the Page access token, plus the ID and username of an Instagram Business account linked to the Page. Hazonlive has no Page selection of its own; we store these details for all Pages you granted in Meta’s dialog. Facebook and Instagram each get a separate entry, and both contain the same Page access token.
What we do not retrieve: Hazonlive does not retrieve or store the name or email address of your Facebook profile. Hazonlive also does not retrieve Page insights, follower data, or posts, comments or messages from other people.
Publishing posts (write): Hazonlive sends Meta the post text with hashtags and, where applicable, one or more images (at most 10, together in one post). We pass images as links that Meta retrieves itself. If an image is not publicly accessible, is in an unsuitable format or is too large, we create a publicly accessible copy of it in our storage for this purpose; AI-generated images are held there at a publicly accessible address anyway. For Instagram, we then query the processing status and the link of the post. Hazonlive does not add a notice of AI generation to the published post; Hazonlive shows the “AI-generated” label only in its own interface.
When posts are published: we publish what you approve or schedule. If you give the AI assistant a time in the chat, it only attaches the time to the draft as a suggestion; the post is scheduled only once you approve it. There is also an optional Autopilot. It is off by default, available only on paid plans, and can only be switched on by owners or administrators of a workspace after a confirmation prompt; a newly connected platform is initially excluded and has to be enabled separately. For the enabled platforms, the Autopilot schedules new posts without your individual approval, at the earliest for the following day, and they are published at the scheduled time. Posts that the AI assistants create in tasks (routines) and posts with a time given in the chat are excluded; they always wait for your approval. Until publishing starts, you can delete or move any scheduled post in the content calendar, or save it as a draft again in the editor. If you switch the Autopilot off, the posts it scheduled whose publishing has not yet started return to approval.
Records of publication: for each post we store the ID that Meta assigns to the published post and the link to it; if publishing fails, we store Meta’s error message (at most 600 characters).
Comments and messages: Hazonlive does not receive or store comments, mentions or direct messages from Facebook or Instagram via Meta’s interface. The app is not registered with Meta for any such notifications and does not request the permissions used to register a Page for them or to manage comments or messages.

Storage and disclosure: access tokens are encrypted with AES-256-GCM before they are stored and cannot be retrieved by your browser. The data described here is stored in our database or file storage at Supabase in Frankfurt; for use, the token is decrypted on our servers at Vercel in Frankfurt. A connection applies to the entire workspace: all members can see the connected Pages and accounts, and members with write access can have posts published through it and can disconnect it. If connecting fails, the error message, which may contain an excerpt of Meta’s response, is sent to Sentry (a US provider; data is stored in its EU data region in Frankfurt am Main) for error diagnostics; data we send to Sentry for error diagnostics and stability monitoring may also contain technical details of requests to Meta. We do not transmit Page or account names, IDs, access tokens, links to published posts or error messages from Meta to AI providers; chat notices about failed publications contain only the beginning of your post text, the channel and a reason worded by Hazonlive. The daily email summary is switched on for the person who completes the setup of a workspace, and can be switched off and on in the settings; it lists scheduled posts with a text excerpt, platform and time, and on Mondays also links to the posts published in the previous week, and it is sent via Resend. Sections 6 and 11 list these service providers, where they are based and the basis for transfers to third countries. Meta (for users in the EU: Meta Platforms Ireland Ltd.) is independently responsible for the processing on Facebook and Instagram themselves, including the published posts. We do not sell data from Facebook and Instagram and do not use it for advertising or to train AI models.

Retention and deletion: we store the connection data until you disconnect it or the workspace is deleted. You disconnect Facebook and Instagram separately under “Integrations” in the settings, using the bin icon; because both entries contain the same Page access token, please disconnect both. If you granted several Pages in Meta’s dialog, Hazonlive shows only one connection per platform; repeat the disconnection until no connection is shown for either Facebook or Instagram. Disconnecting deletes the respective entry, including the token, from our database immediately, but does not revoke the permission at Meta; the entry remains in the daily database backups until they are routinely overwritten (see section 9). IDs and links of published posts and error messages from Meta remain stored until you delete the post in Hazonlive or the workspace is deleted; deleting the post removes them from our database immediately. There is currently no fixed deletion period for them. Image copies made specifically for publishing remain until the workspace is deleted. Chat notices about failed publications are deleted with the respective chat history, by default 90 days after the last activity in it; because new notices appear in the same chat, each new notice extends the retention of the older ones, and pinned chats are exempt from this period; chats in the trash are deleted 90 days after being moved there (see section 9). Log entries about connecting, disconnecting and publishing are deleted after 12 months. Deleting a post in Hazonlive does not remove it from Facebook or Instagram. If you delete your user account, the workspaces you own are deleted with all connections, posts and log entries; a connection you set up in a workspace owned by another person remains there until it is disconnected there. Independently of this, you can revoke Hazonlive’s access in the settings of your Facebook account; we are not notified of this, and the data stored with us remains until it is deleted in one of the ways described. All ways to request deletion, including by email, are described on our page Delete your data.

5. Data storage

In our database, master data and usage data are stored on servers in the European Union:

Database: Supabase (AWS eu-central-1, Frankfurt)

The application is hosted by Vercel; its serverless functions are configured in the Frankfurt region (fra1), while Vercel’s primary processing facilities are in the USA. Transactional emails (login, password reset, notifications) are sent by Resend via AWS eu-west-1 (Ireland); Resend stores the data, including message content, in the USA.

AI prompts and image descriptions are transmitted for processing to the AI providers in the USA named in section 6 (Standard Contractual Clauses pursuant to Article 46 GDPR). Payment data is processed by the payment service provider Stripe. Vercel and Resend are based in the USA, Supabase in Singapore; for transfers to third countries, see section 11.

5a. Data security

We protect personal data — in particular sensitive content from connected Google services (Gmail, Google Calendar) — through technical and organisational measures pursuant to Article 32 GDPR:

Encryption in transit: All connections to Hazonlive and between our services are exclusively TLS-encrypted (HTTPS).
Encryption at rest: Data is stored encrypted in the database (encryption at rest, AWS eu-central-1, Frankfurt). Access and refresh tokens for connected mailboxes and social media channels are additionally encrypted at the application level using AES-256-GCM.
Access control and tenant isolation: Access to data is strictly separated by workspace and enforced at the database level through Row Level Security (RLS). Access to your data by our staff occurs only to the extent necessary for operation, support or legal obligations (principle of least privilege).
Operator support access: For troubleshooting we can open a time-limited view of your workspace ("view as customer"). This access is bound to a cryptographically signed session proof, requires two-factor authentication, expires automatically after 30 minutes (read mode) or 15 minutes (edit mode), and is logged. No email is sent while it is active.
Content from connected mailboxes is blocked by default in this mode — message bodies, reply drafts and writing samples learned from your sent folder. It becomes visible only if you explicitly release access in the settings; that release is capped at eight hours, revocable at any time, and appears in your audit log. For Gmail mailboxes the block remains in place even with your release: Google's rules for access to mailbox data provide for no exception based on your consent.
Limited use of Google data: Content from Google services is used solely for the functions visible to you described in section 4a, is not disclosed to third parties (other than the processors listed in section 6 for the purpose of providing these functions) and is not used for advertising or to train AI models. This complies with the Google API Services User Data Policy, including the Limited Use requirements.
Data minimisation: Only the data required for the respective function is transmitted to AI providers (e.g. for the inbox the sender, subject line and shortened message text, no attachments (exception: text taken from the attachments of invoices if “Read invoices in detail” is switched on)).
Operations and monitoring: We use access logging and error/stability monitoring to detect security incidents. You can revoke mailbox connections at any time in the settings; revoked tokens become invalid and the associated connection is removed.

6. Processors and disclosure to third parties

We use the following service providers as processors (Article 28 GDPR) or as independent controllers:

Anthropic PBC (Claude)

Purpose: text AI, AI agents, document analysis, conversation processing, checking and describing images, still frames from uploaded Reel videos and the transcript of what is spoken in them, from which the assistant suggests text and a cover image, and web search for the assistants (e.g. lead search, trend research)
Data transferred: user prompts, conversation histories, images (including still frames from uploaded Reel videos) and documents for checking, description and text recognition; for Reels, the transcript of what is spoken (text, not the audio track); for web search, the search term the assistant derives from your task. Anthropic runs the search with Brave Search as its own sub-processor. Search terms containing direct identifiers (email address, phone number, IBAN, card or tax number) are blocked technically
Server location: USA (SCCs pursuant to Article 46 GDPR)
Privacy: https://www.anthropic.com/privacy

OpenAI OpCo, LLC (text AI, semantic search and speech recognition)

Purpose: text AI as a fallback provider when Anthropic is briefly unavailable, and for assistants for which OpenAI is selected in your workspace; semantic search so that the assistants find relevant content (embeddings); speech recognition: the audio track of uploaded Reel videos is converted into text so that the assistant can write the post text. We do not store the still frames, the audio track or the transcript; a short summary of the video remains on the post so that the text can be rewritten. The video itself is kept in your workspace's file storage (Supabase, Frankfurt) under a non-guessable, publicly accessible address so that Instagram can load it when publishing
Data transferred: in the fallback or selection case, user prompts and conversation histories; for search, text passages from your knowledge base, summaries of chat conversations, texts of approved social media posts and the assistants' search queries; for speech recognition, the audio track of uploaded Reel videos (may contain voices and spoken content)
Server location: USA (SCCs pursuant to Article 46 GDPR)
Privacy: https://openai.com/policies/privacy-policy

OpenAI OpCo, LLC (image generation)

Purpose: AI image generation on user request
Data transferred: image prompts (text descriptions); when you have an existing image revised, that image; when an image is to show a person from your media library, the photos you marked as a person reference for this
Server location: USA (SCCs pursuant to Article 46 GDPR)
Privacy: https://openai.com/policies/privacy-policy

Google LLC (Gemini API)

Purpose: backup image model — used only if the primary image model (OpenAI) fails when generating a new image
Data transferred: image prompts (text descriptions) and up to three images most recently generated by AI in your workspace as a style reference
Server location: USA (SCCs pursuant to Article 46 GDPR)
Privacy: https://policies.google.com/privacy

Supabase Pte. Ltd. (Singapore)

Purpose: database, authentication and file storage
Server location: AWS eu-central-1, Frankfurt am Main; because the contracting party is based in Singapore, Standard Contractual Clauses (Article 46 GDPR) apply in addition
Privacy: https://supabase.com/privacy

Vercel Inc.

Purpose: hosting and delivery of the application
Server location: USA (the provider’s primary processing facilities); the application’s serverless functions are configured in Frankfurt (fra1). The transfer is based on the provider’s certification under the EU-U.S. Data Privacy Framework (adequacy decision, Article 45 GDPR); in addition, its data processing addendum (effective 31 March 2026, incorporated via section 10.1 of its terms of service) contains Standard Contractual Clauses (Article 46(2)(c) GDPR)
Privacy: https://vercel.com/legal/privacy-policy

Plus Five Five, Inc. (Resend)

Purpose: sending transactional emails (sign-up confirmation, password reset, system notifications, and call notifications if the telephone feature is activated)
Data transferred: email address, user name, mail content; for call notifications also the caller’s name and phone number, the AI summary of the call (including urgency and action items) and the note
Server location: USA (storage, including message content and delivery logs); emails are sent via AWS eu-west-1, Ireland. The transfer is based on the Standard Contractual Clauses in its data processing addendum (Article 46 GDPR), which alone govern transfers from Switzerland; in addition, the provider is certified under the EU-U.S. Data Privacy Framework (Article 45 GDPR; re-certification currently under review)
Privacy: https://resend.com/legal/privacy-policy

Functional Software, Inc. (Sentry)

Purpose: error diagnostics and stability monitoring of the application on the basis of Article 6(1)(f) GDPR (see section 8). With your consent (cookie banner: „Alle akzeptieren“ / “Accept all”) additionally session replay for error analysis — text inputs are masked and media blocked in the process.
Data transferred: technical error and performance data (browser, device, error message); for errors on our servers also details of the request concerned, including the IP address, and error messages that may contain excerpts of responses from connected services; from your browser, with consent, additionally IP address and masked session recordings
Server location: Sentry’s EU data region in Frankfurt am Main; because the provider is based in the USA, Standard Contractual Clauses (Article 46 GDPR) apply in addition
Privacy: https://sentry.io/privacy/

Twilio Inc.

Purpose: telephone reception (incoming calls, call notes) — only relevant if the telephone feature is activated in your workspace
Data transferred: caller’s phone number, call metadata and call content — the caller’s speech is processed by Twilio for speech recognition and responses are spoken via text-to-speech; calls are not recorded
Server location: USA; the transfer is based primarily on the provider’s certification under the EU-U.S. Data Privacy Framework (Article 45 GDPR), alternatively on its Binding Corporate Rules (BCRs) or Standard Contractual Clauses (Article 46 GDPR) under its data processing addendum
Role: Twilio acts as a processor for providing the telephone service and as an independent controller for its own purposes (including operating as a communications provider, security and abuse detection, legal obligations, product improvement); this concerns connection and usage data and, to that extent, also call content
Privacy: https://www.twilio.com/legal/privacy

Browser vendors’ push services (web push)

Purpose: delivering push notifications to your browser — only if you choose the “Push” or “Both” channel under Settings → Notifications and your browser allows notifications. Currently these are alerts about pending approvals (at most once per hour per workspace) and, for owners and admins, payment reminders (at most once a day per workspace).
Stored by us: for each device, the push address generated by your browser and two encryption keys, linked to your account and workspace, plus the time of registration or of the last message the push service accepted.
Data transferred: the push address, the time, our sender identification (VAPID) and the encrypted content of the message. The push service cannot read the content; it only contains the number of pending items or upcoming payments, a fixed text and a link into the app, never names, amounts, post texts or message content. If your device cannot be reached, the push service holds the encrypted message for up to four weeks.
Recipient: your browser’s push service — Google, Mozilla, Apple or Microsoft (e.g. Google for Chrome, Mozilla for Firefox, Apple for Safari). Your browser determines which one, not us; we do not send to any other push service.
Role: the push services are not processors. We have no contract with them for delivering push notifications; they operate the service for your browser under their own responsibility, comparable to the provider of your email inbox.
Location: the providers are headquartered in the USA and may process data there. The transfer is necessary for the delivery you switched on (Article 49(1)(b) GDPR); the content is end-to-end encrypted.
Switching off: under Settings → Notifications, choose the “Email” channel (it applies to approval alerts and payment reminders), turn off “Instant on approvals” and “Payment reminders”, or revoke notifications in your browser settings. How long we keep the registration is set out in section 9.

Stripe (payment service provider)

Purpose: technical payment processing (card payment, SEPA direct debit and others) on behalf of the provider
Data transferred: name, email address, billing address, payment data (PCI-DSS-compliant via Stripe)
Provider: Stripe Payments Canada, Ltd. (master agreement) and Stripe, LLC (processing), USA
Role: Stripe acts as a processor for payment processing and as an independent controller for its own purposes (including towards banks and payment method providers, fraud and risk checks, anti-money laundering and identity checks, legal obligations, product improvement)
Server location: among others the USA; the transfer is based primarily on Stripe, LLC’s certification under the EU-U.S. Data Privacy Framework (Article 45 GDPR), alternatively on Standard Contractual Clauses (Article 46 GDPR)
Personal data may be transferred, processed and stored outside of Canada and, as a result, may be subject to disclosure as required by applicable law
Privacy: https://stripe.com/privacy
Note: The purchase contract for paid plans is concluded with the provider (AGAPE INNOVATIONS GROUP INC.); Stripe handles the payment technically. For details, see Terms and Legal notice (Imprint).

7. Your rights

You have the right to:

Access to your stored data (Article 15 GDPR)
Rectification of inaccurate data (Article 16 GDPR)
Erasure of your data (Article 17 GDPR)
Restriction of processing (Article 18 GDPR)
Data portability (Article 20 GDPR)
Objection to processing (Article 21 GDPR)
Withdrawal of consent given (Article 7(3) GDPR)

To exercise your rights, please contact us at: kontakt@hazonlive.com

8. Cookies and consent

We use technically necessary cookies for authentication. Error diagnostics via Sentry (see section 6) are based on Article 6(1)(f) GDPR (legitimate interest in stable, error-free operation) and do not depend on your consent: we report errors on our servers as well as technical error and performance data from your browser. Only with your consent (“Accept all” in the cookie banner) do error diagnostics in your browser additionally transmit your IP address and further request data, and do we enable session replay (text inputs are masked, media blocked). We do not use marketing or advertising cookies. You can change your choice at any time via “Cookie settings” in the footer.

9. Retention period

Personal data is deleted as soon as the purpose of storage no longer applies. If you delete your user account, we delete it immediately together with the workspaces you own and all data stored in them. In workspaces owned by other people, content and log entries you created may remain without being linked to your account; for connections to Facebook and Instagram, see the last paragraph of this section.

We automatically delete chat histories with the agents, including attached images, 90 days after the last activity in them. Chats you move to the trash are permanently deleted 90 days after being moved; until then you can restore them or delete them permanently at any time. Pinned chats are exempt from automatic deletion as long as they are pinned and not in the trash.

Backups: our hosting provider Supabase backs up the database automatically every day. Data we delete from the database remains in these backups until they are routinely overwritten; on our current plan, Supabase keeps the backups of the last 7 days. Files from our file storage (e.g. images) are not included in these database backups.

Results of the legal assistant (contract traffic light, GDPR check, DPA draft, legal question) are stored as a history so you can review them again. Only the structured results and metadata are kept, never the full contract or input text. They are deleted automatically after 90 days, and every entry can also be deleted manually at any time.

Entries in the invoice overview (invoice data recognised from invoice emails, including their history) are deleted 24 months after they are closed, open ones at the latest 36 months after they were recognised, and immediately when you disconnect the last connected mailbox of that provider. Details in section 4a.

We keep your notification settings (which notifications are on, which channel, when the last one was sent) until you delete your account, leave the workspace or are removed from it, or the workspace is deleted. If you leave a workspace or are removed from it, we delete these settings immediately, at the latest the following night, and you no longer receive notifications from that workspace. We delete push registrations of your devices by the following day at the latest once you no longer use push in any of your workspaces; also as soon as the push service reports a registration as invalid, and at the latest 12 months after the last registration or the last message to the device that the push service accepted. Details in section 6 under “Browser vendors’ push services”.

We store connection data for Facebook and Instagram (IDs, names, encrypted access tokens) until you disconnect the respective connection or the workspace is deleted. IDs and links of published posts and error messages from Meta remain stored until you delete the post in Hazonlive or the workspace is deleted; deleting the post removes them from our database immediately. There is currently no fixed deletion period for them. Image copies made specifically for publishing remain until the workspace is deleted. Log entries about connecting, disconnecting and publishing are deleted after 12 months. If you delete your user account, connections you set up in a workspace owned by another person remain there until they are disconnected there. Details in section 4c.

10. Right to lodge a complaint

You have the right to lodge a complaint with a data protection supervisory authority.

11. Transfer of data to third countries

Personal data is transferred to the USA in the context of AI processing (Anthropic, OpenAI, Google as a backup image model), telephone reception (Twilio) and payment processing (Stripe); this includes the search terms of the web search, which runs via Anthropic. For Twilio and Stripe, the transfer is based primarily on their certification under the EU-U.S. Data Privacy Framework (Article 45 GDPR), alternatively on the mechanisms named in section 6. For error diagnostics (Sentry), the provider is a company based in the USA; the data is stored in Sentry’s EU data region in Frankfurt am Main. The contracting party for database and file storage (Supabase) is a company based in Singapore; the data is stored in Frankfurt am Main. The other transfers are based on the Standard Contractual Clauses of the European Commission pursuant to Article 46(2)(c) GDPR, together with supplementary technical and organisational measures of the respective providers.

Hosting (Vercel) and email delivery (Resend) are provided by companies based in the USA. The application’s serverless functions are configured in the Frankfurt am Main region, while Vercel’s primary processing facilities are in the USA; emails are sent via the Ireland region, and Resend stores the data in the USA. Transfers to Vercel are based on its certification under the EU-U.S. Data Privacy Framework and the European Commission’s adequacy decision for it (Article 45 GDPR) and, in addition, on the Standard Contractual Clauses in its data processing addendum (Article 46(2)(c) GDPR). Transfers to Resend are based on the Standard Contractual Clauses in its data processing addendum (Article 46(2)(c) GDPR), which alone apply to transfers from Switzerland; in addition, Resend is certified under the EU-U.S. Data Privacy Framework (Article 45 GDPR; re-certification currently under review).

If you switch on push notifications, we deliver them via your browser’s push service (Google, Mozilla, Apple or Microsoft, each headquartered in the USA). The basis is Article 49(1)(b) GDPR; details in section 6.

In addition, the operator AGAPE INNOVATIONS GROUP INC., based in Toronto, Canada, has access to personal data as the controller. This transfer is based on the adequacy decision of the European Commission for Canada (Article 45 GDPR), which applies to commercial organisations subject to the Canadian data protection act PIPEDA.

As of: 23 October 2026.