This Privacy Policy explains how Ben Hadad, trading as "AssistantLabs" ("we", "us", "our"), collects, uses, stores, shares and protects personal data when you use our cloud-based AI assistant platform (the "Service"). The Service enables businesses to build and operate AI-powered assistants across WhatsApp Business, Instagram Direct, Facebook Messenger, website chat and LinkedIn, with an in-app inbox, CRM, analytics and optional integrations. Customers may also connect an external AI assistant, such as Claude or ChatGPT, to their workspace through our optional MCP connectors.
We are committed to protecting your privacy and to complying with the Israeli Privacy Protection Law, 5741-1981 (including Amendment No. 13, in force since August 2025) and the regulations made under it, the EU General Data Protection Regulation (GDPR) where applicable, and other relevant data-protection legislation.
The controller for account, billing and platform data processed through the Service is:
Ben Hadad / AssistantLabs
Address: Ha'Aniya Arinpura 8, Netanya, Israel
Email: legal@assistantlabs.io
When a Customer (a business) uses AssistantLabs to interact with its own end-users, the Customer is the data controller for end-user personal data and AssistantLabs acts as the data processor on the Customer's behalf, processing that data only on the Customer's documented instructions and to provide the Service.
We collect and process the following categories of personal data:
| Field | Description |
|---|---|
| id | Unique user identifier |
| displayName | User's display name |
| Email address | |
| lastLoginAt | Timestamp of last login |
| profilePicture | Profile photo URL (optional) |
| settings.language | Preferred language |
| Field | Description |
|---|---|
| cardHolderName | Name on credit card |
| cardHolderId | ID document number of cardholder (Israeli invoicing requirement) |
| cardType | Card brand (Visa, Mastercard, etc.) |
| lastFourDigits | Last 4 digits of card number |
| expMonth / expYear | Card expiration date |
| token | Tokenized card reference (no full card numbers stored by us) |
| businessName | Company or business name for invoicing |
| businessId | Israeli business ID or personal ID for tax |
| receiptEmail | Email for receipts and invoices |
| Standing order & usage records | Subscription and usage-charge history for billing |
Business information, FAQs, product catalogs, links, guidelines, scenarios, persona settings and other content that Customers provide (directly or via the Logos AI configuration assistant) to configure their AI assistants. Customers may also import content by allowing us to scan a website they provide.
| Field | Description |
|---|---|
| firstName / lastName | Contact name |
| phone | Phone number |
| Email address | |
| notes | Free-form notes about the contact |
| customFields | Custom key-value data |
| labels | Categorization labels |
| source | Origin channel (WhatsApp, Instagram, Messenger, Web, LinkedIn, Wix, Monday.com, Shopify, WooCommerce, import, manual or API) |
| consentStatus | Data-processing consent (granted, denied, pending, unknown) |
| marketingConsent / optedOut | Opt-in status and opt-out flag/timestamp for marketing communications |
| External IDs & synced orders | Identifiers and order data from connected integrations (e.g., wixContactId, mondayItemId, WooCommerce/Shopify customer or order records) |
Messages exchanged between end-users, AI assistants and the Customer's human agents, including text and media content, timestamps, message direction, sender type, read/reaction status, channel metadata and platform message identifiers.
When a Customer connects their own Meta ad account to the Service through Facebook Login for Business, we receive and store, for that Customer only: the access token Meta issues, the identifiers and names of the ad accounts, Facebook Pages and business portfolios the Customer grants us, and the campaigns, ad sets, ads and delivery metrics (impressions, clicks, spend, results) for advertising our Service creates or manages in that ad account. We use this data solely to create, edit, pause and report on advertising on the Customer's instruction. We do not use it to build profiles of end-users, and we do not receive the personal data of the people who see or click those ads. A Customer can disconnect the ad account at any time in the Service, which deletes the stored token and the connection.
When a Customer connects an Instagram professional account and creates a comment automation (for example, "when someone comments a keyword on this post, reply under the comment and send them a private message"), we receive, for that Customer's account only: the access token Instagram issues, the account's identifier and username, and, for comments on the account's own posts and mentions of the account, the comment text, the comment and post identifiers, and the commenter's Instagram-scoped identifier and username. We use this data solely to check each comment against the keywords the Customer chose, to post the Customer's public reply and private message, and to show the Customer an activity log of what was sent. We never reply to the account's own comments, we reply at most once per comment, and we do not use comment data for advertising, profiling or any purpose other than running the automations the Customer created. If the Customer removes our app in Instagram's settings, we stop all automations on that account and delete its access token; if the Customer requests deletion of their data through Instagram, we delete the connection, its automations and its activity log.
Conversation counts, automated-vs-human classification, message volumes, response metrics, session data, IP addresses, browser type, device information and usage patterns, collected through Firebase Analytics and exported to BigQuery in aggregated form for analytics and billing.
When a Customer uses Booking: its services and opening hours, and each appointment's service, date and time, duration, price, status and source, together with the customer's name, phone number, email address and any notes. A stay also records its arrival and departure dates and party size.
The people on a Customer's Leads board and in its customer groups and sales journeys: name, phone number, email address, lead stage and temperature, the lead's owner, its next step and when that is due, when a person last spoke to them, why a lead was lost, the outcomes and notes logged against each lead, how far each person is through a journey and what they were sent, and order or abandoned-cart details from a connected store.
Tasks on a workspace's board: title, contents, notes, status, assignees, project and dates, and any link to the customer or conversation a task came from.
Tickets opened from customer conversations: state, priority, labels, topic, summary, what a person needs to do, the customer's name, the linked conversation and its messages, the assignee and the team's private notes. Helpdesk also keeps its own contact records with their notes, and an audit log of actions taken.
When a Customer uses Finance: its income and expense entries; invoices, receipts, credit notes and payment requests, with the client's name, document number, date, status, amount and currency; and the payments received against them. When a Customer connects its own Morning (Green Invoice) account, we also read its clients from Morning: name, one email address, one phone number, a masked tax ID and the amounts invoiced, paid and still open, with their latest documents. Morning clients are read live and are not stored by us; documents and expenses imported from Morning are stored in Finance until the Customer disconnects Morning, which removes them.
Email and WhatsApp campaigns a Customer creates: their names, content, schedule and status, summary numbers (sent, delivered, opened, clicked, unsubscribed), and for each recipient their name, email address or phone number and what happened with each message. Email templates and each person's email history. Social posts: captions, media, target accounts, publishing status and per-post performance.
People who fill in a form on a Customer's landing page: name, phone number, email address, any other form fields, the page, whether they agreed to marketing messages, the referrer and campaign attribution, and whether the lead reached the CRM and received its follow-up.
Marketing email sent through the Service records, per recipient, whether and when each message was opened, how many times, which link was last clicked and when, and any unsubscribe. See "Email open and click tracking" in the Cookies & Analytics section below.
Each contact has a timeline of what has happened with them: notes, label changes, opt-outs and opt-ins, changes to their contact fields, conversation status and assignment changes, flags, when the AI assistant was paused or switched back on, order totals, and calls and other events sent in from the Customer's own systems.
Companies a Customer keeps in its CRM: name, website, industry, description, relationship to the business and labels, with the people linked to each company by name and job title.
The people who work in a Customer's workspace: their names and roles, who is assigned to which ticket or task, and who owns which lead. Each person also has an internal identifier. The ChatGPT connector does not return it, and tickets are assigned there by name.
What a Customer asks a connected AI assistant to save is stored in the workspace in the same way as data entered in the app. For example, adding a contact or a lead stores the name, phone number, email address and note given in the chat, booking an appointment stores the customer's details and notes, updating a contact stores the change, and updating a lead stores its owner, next step, last contact date, lost reason or note. See the AI Assistant Connectors (MCP) section below.
| Legal Basis | Processing Activity |
|---|---|
| Performance of contract | Providing the Service, processing payments, managing accounts, delivering AI assistant functionality |
| Legitimate interest | Service improvement (using aggregated/de-identified data), analytics, security monitoring, fraud prevention, debugging |
| Consent | Marketing communications, optional analytics, cookie placement |
| Legal obligation | Tax record keeping, responding to lawful government requests, maintaining billing records |
We do not sell or rent personal data, and we do not use identifiable Customer Content or end-user conversations to train or fine-tune general-purpose AI models. Only aggregated and anonymized data is used to improve the Service.
We share personal data with the following sub-processors, strictly as necessary to provide the Service. Integration providers receive data only when the Customer connects the relevant integration.
| Sub-Processor | Purpose | Location |
|---|---|---|
| Google Cloud Platform (GCP) | Cloud infrastructure, Firestore database, Cloud Run, BigQuery analytics, Cloud Tasks, storage | Iowa, USA |
| Firebase (Google) | Authentication, hosting, analytics, Cloud Functions | USA |
| OpenAI | AI language-model processing for assistant responses (see the note below on ChatGPT, which is separate) | USA |
| Anthropic | AI language-model processing for the Logos configuration assistant | USA |
| Google (Gemini / Generative AI) | AI language-model processing | USA |
| Meta Platforms Ireland Ltd | WhatsApp Cloud API, Instagram Messaging API, Messenger Platform, Marketing API (advertising created and managed in a Customer's own ad account) | Ireland / USA |
| Meta Platforms Ireland Ltd (Meta Pixel) | Measurement and advertising analytics on our own dashboard/marketing web pages (account users only; not applied to end-user conversations). Placed subject to consent where required. | Ireland / USA |
| Resend | Sending and receiving email, including message content: account notifications, workspace invitations, onboarding and support mail; mail an AI assistant sends and receives on a Customer's email channel; and a Customer's campaign, journey and booking email, unless the Customer connected its own Gmail or Microsoft 365 mailbox for sending | USA |
| Google (Gmail API) | Sending and receiving a Customer's email-channel mail and organisation notification email, where the Customer connects a Gmail account | USA |
| Microsoft (Microsoft Graph) | Sending and receiving a Customer's email-channel mail, where the Customer connects a Microsoft 365 mailbox | USA |
| Flashy | WhatsApp template / marketing-message delivery (when used) | Israel |
| Tranzila (InterSpace Ltd) | Payment processing, credit-card tokenization, standing orders | Israel |
| Morning (Green Invoice) | Reading clients, invoices, receipts, credit notes, expenses and payment records from the Customer's own Morning account into Finance, where the Customer connects it with its own API key | Israel |
| Customer-enabled integrations | Wix, Monday.com, Shopify, WooCommerce, Fireberry, Calendly, Google Calendar/Sheets — contact, order, scheduling and CRM synchronization | Varies by provider |
A current sub-processor list is available on request at legal@assistantlabs.io.
An external AI assistant that a Customer connects through an MCP connector is not a sub-processor: it receives data on the Customer's own instruction, under the Customer's own agreement with that AI provider. That holds for ChatGPT too: OpenAI is listed above for generating our own assistants' responses, which is a separate arrangement from a Customer connecting ChatGPT. See the AI Assistant Connectors (MCP) section below.
We publish remote MCP (Model Context Protocol) connectors that let a Customer connect an external AI assistant — such as Claude or ChatGPT — to their own AssistantLabs workspace and then ask it about their business in plain language. Connecting one is optional and entirely Customer-initiated. If a Customer never connects a connector, nothing in this section applies to them.
A single connector address — /mcp — can carry every AssistantLabs product: the assistant and its conversations, the task board, the CRM, Sales, Booking, Helpdesk, Marketing, Landing Pages and Finance. Which of them a given connection actually reaches is decided by the permissions the Customer approves, not by the address. A connection approved only for customer replies loads only those tools, and a product the Customer did not approve is never loaded at all. Several products also keep a dedicated address, for a Customer who prefers to connect one product on its own, and ChatGPT has its own address with a smaller, fixed set of tools. Everything is served from https://mcp.assistantlabs.io.
| Product | Dedicated address | What it can reach |
|---|---|---|
| Assistant & conversations | /mcp — also the combined address (alias /insights/mcp) | Assistant configuration and knowledge, conversations and their messages, contacts, segments, channels and WhatsApp templates, follow-ups, business memory, owner approvals and notifications, custom integrations |
| Tasks | /tasks/mcp | The workspace task board — tasks, their contents and views |
| CRM | /crm/mcp | CRM contacts, companies and custom records with their history and attached files, CRM tasks, automations and their run logs, reports and dashboards, roles and team members, webhooks, entity schemas, record page layouts |
| Sales | /sales/mcp | Leads, customer groups, journeys and their audiences, participants, journey statistics, WhatsApp templates |
| Booking | /booking/mcp | Services, opening hours, free slots, appointments and their customers, booking messages |
| Helpdesk | /mcp | Tickets, their messages and team notes, helpdesk contacts, team members, workflows, settings and the audit log |
| Marketing | /marketing/mcp | Marketing channels, social posts and performance, strategy, research and product profile, paid ads, email and WhatsApp campaigns with their recipients, email templates and history, Instagram posts, comments and automations, media library |
| Landing Pages | /mcp | Landing pages and the leads their forms collected |
| Finance | /mcp | Read only: money summary, open invoices, payments received, Morning clients and their documents |
| ChatGPT | /chatgpt/mcp | The fixed set of tools offered to ChatGPT, across the products above. It leaves out custom CRM records and attached files, and companies are read only. Lists show customer names, who wrote, when and whether the customer is waiting, without message text; a phone number or email address appears only when the Customer asks for that person's details. Team members' internal identifiers are not returned. Phone numbers, email addresses, card numbers and ID numbers in common formats are automatically hidden inside message text |
Each product is resolved independently. The consent screen shows which products a connection includes. A connection to the combined address always includes the task board; apart from that, approving one product does not require approving another, and on the ChatGPT connector the task board is optional too. If one product's service is unavailable, that costs only its own tools, not the rest of the connection.
A connector reads data that already exists in the Customer's workspace, when the connected AI assistant calls a tool in response to something the Customer asked. It also stores what the Customer gives it in the chat: tools such as add_contact, update_contact, add_lead and book_appointment save the names, phone numbers, email addresses, notes and other details provided, in the workspace (see Data Added Through a Connector above). Depending on which connectors are authorized and which permissions are granted, the data returned may include:
This data belongs to the Customer. Where it contains personal data about the Customer's end-users, the Customer remains the controller and AssistantLabs remains the processor, exactly as described above — connecting an AI assistant does not change those roles.
A connector reaches nothing until the Customer authorizes it, in one of two ways:
Permissions are enforced, not advisory. Tools are exposed according to the permissions the credential actually holds, and every request is authorized again by the underlying API. A tool whose permission was not granted is not offered, and a call to it is refused. Granting read access to conversations does not grant the ability to send messages, change assistant configuration or edit records — those are separate permissions, and the Customer can leave any of them off on the consent screen. A connection that includes the task board can also change tasks.
On the ChatGPT connector, the consent screen offers only the permissions its tools can use. Sending messages to customers, changing contacts, booking and changing appointments, publishing posts, pausing ads and changing helpdesk tickets start switched off, and the Customer switches each one on itself. Changing leads and the assistant's configuration start switched on, and the Customer can switch them off. Where ChatGPT asks for it, it can also read the email address of the person who approved the connection, so that a workspace admin can keep a company address from connecting outside the company's workspace. The consent screen says so when this applies, and this reveals only that address and an account identifier.
Where a Customer does grant write permissions, the connected AI assistant can act in the workspace on the Customer's instruction: send a message to a customer, add or update a contact, add a lead, book or cancel an appointment, update a CRM record (not through ChatGPT), change assistant configuration, or publish a post. Booking and cancelling can notify the customer. Moving a lead or logging an outcome runs what the Customer attached to that stage, and the tool reports what ran, except on the ChatGPT connector, which declines a move or outcome that would run attached steps so that it is done in the app. On the ChatGPT connector, adding a lead, updating a lead and logging an outcome send nothing, and booking or completing an appointment updates a linked lead's stage without running what the Customer attached to that stage. The Customer is responsible for the instructions it gives and for the permissions it approves.
The connector service has no database of its own and writes nothing to disk: each tool call becomes an authenticated request to the AssistantLabs API, and the response is handed back to the AI assistant that asked for it.
It does hold some data briefly in memory. When an analysis tool reads a conversation's messages, for example to search conversations or calculate conversation statistics, the connector keeps those messages so the next tool call does not have to read them again. Each entry is held for at most 10 minutes, is tied to the credential that read it and is never served to another credential, and is replaced as soon as the conversation changes. The connector holds at most 20,000 messages at a time across all connections, with the oldest giving way first. This cache is never written to disk and is lost when the service restarts. The connector also remembers which workspace a credential belongs to for 30 seconds, so that it does not have to look this up on every request.
Credentials are never stored by the connector and never written to logs in full — an API key appears in operational logs only as a masked fragment. We keep ordinary operational logs (timestamps, request paths, response codes, tool names, the workspace and connecting client, and error messages) for 30 days, for reliability, security monitoring and abuse prevention. Every tool call and API request also writes a usage record: the time, the workspace, the credential and connecting client, the tool or route, the result and how long it took. A usage record does not contain the content of the request or the response. Usage records sit in those logs, and they may also be copied to our own analytics store in Google BigQuery for billing, usage reporting and abuse prevention, where no fixed deletion period applies. Logs and usage records are not used to profile end-users. Data reached through a connector is not used to train or fine-tune any AI model.
Responses are bounded, so that a broad question cannot quietly drain an entire history: a single tool result is capped at 60,000 characters, an analysis call scans at most 300 conversations, and at most 200 messages are read per conversation.
This is the most important thing to understand before connecting one.
When a Customer connects an external AI assistant, the data a tool returns goes to the assistant the Customer connected — for Claude, to Anthropic; for ChatGPT or Codex, to OpenAI. That is inherent in the request: the assistant cannot answer a question about the Customer's conversations without receiving them. AssistantLabs does not select that provider and does not control what happens to the data once it arrives. Its handling, retention and any human review are governed by the Customer's own agreement with that AI provider and by that provider's privacy policy — not by this one.
We recommend that Customers read the AI provider's terms before connecting, and authorize no permission broader than they need. We do not sell or rent this data, and we disclose it to no recipient other than the AI provider the Customer chose to connect and the infrastructure sub-processors listed above.
Where a Customer is a controller under the GDPR or Israeli privacy law and its end-users' personal data will reach an AI provider through a connector, the Customer is responsible for having a lawful basis for that disclosure and for reflecting it in its own privacy notice.
The connector stores no workspace content itself: what a tool saves goes into the workspace. Conversation messages may be held in memory for up to 10 minutes, as described above, and then expire on their own. Operational logs are retained for 30 days, and usage records may be kept longer, as described above. The underlying workspace data keeps the retention periods set out in the Data Retention section below. Data delivered to a connected AI provider is retained under that provider's policy, and only the Customer can delete it there.
A Customer may disconnect at any time — under Connected apps in AssistantLabs, or by deleting the developer API key the connector uses. Access stops immediately. The Customer should also remove the connector in its AI assistant, because removing it there does not revoke the authorization it holds. Messages still in the in-memory cache can no longer be reached by that connection and expire within 10 minutes, operational logs age out after 30 days, and usage records are kept as described above. Anything already delivered into an AI assistant's chat history must be removed in that assistant.
For questions, access requests or complaints specifically about the MCP connectors, contact legal@assistantlabs.io, or use the full contact details in the Contact Us section below. We respond within 30 days.
The MCP connectors are built for ordinary business correspondence: enquiries, orders, bookings, support and the records a business keeps about its own customers. They are not intended for, and must not be used to process, regulated health data — including Protected Health Information as defined by the US HIPAA Privacy Rule (45 C.F.R. § 160.103) — or payment card data governed by PCI DSS. No connector tool can make a payment, transfer money or charge a card, and our billing stores no full card numbers. No tool is built to read medical records, but a connector reads what a workspace holds: if a customer or a team member wrote health information or a card number into a message, note or ticket, a tool that reads that item can return it. On the ChatGPT connector, card numbers in common formats are automatically hidden, but health information is not. Do not use a connector for health records. Finance tools read invoices and payment records (client names, document numbers, dates and amounts) and cannot issue, change or send an invoice. A WhatsApp template sent through a connector, including ChatGPT, is billed to the Customer by WhatsApp. Two further capabilities can have a financial effect, and only where a Customer grants them outside the ChatGPT connector: a helpdesk workflow the Customer builds with a refund step refunds that order through the Customer's connected Shopify store, and advertising tools can switch on a paid ad, which spends that ad account's budget with Meta. A Customer who is a healthcare provider, or otherwise handles special-category health data, should not connect an assistant that carries that data to a third-party AI assistant through these connectors. A Customer who does so is responsible for that decision and for any consent or legal basis it requires. This restriction is in addition to, not instead of, the terms of the AI provider whose assistant is connected.
Personal data is transferred to and processed in the United States (Google Cloud Platform, Iowa region; OpenAI; Anthropic; Google, including the Gmail API and Meta Pixel measurement; Resend; Microsoft, where a Customer connects a Microsoft 365 mailbox), Ireland/USA (Meta), Israel (Tranzila, Flashy, Morning) and other locations depending on the integrations a Customer enables. These transfers are necessary to perform our contract and are safeguarded by:
We do not apply a default deletion period to the data a Customer holds in its workspace: it stays until the Customer deletes it or asks us to. The periods below are the ones that apply.
| Data Category | Retention Period |
|---|---|
| Account data | Until the account is deleted, which removes it immediately (see How to delete your data). An account closed any other way is removed within 30 days of termination |
| Conversation data | Until deleted by Customer via dashboard; no default deletion period. Deleting an account immediately removes the conversations of the assistants it owns personally; a workspace's conversations are removed within 30 days of a written request |
| CRM contacts | Until deleted by Customer; no default deletion period. Deleting an account immediately removes the contacts of the assistants it owns personally; a workspace's contacts are removed within 30 days of a written request |
| Appointments, leads, tasks, helpdesk tickets, landing-page leads, campaigns (with their open and click data) and social posts | Until deleted by Customer or on written request; no default deletion period; removed within 30 days of a written request |
| Finance data | Until deleted by Customer; documents imported from Morning are removed when the Customer disconnects Morning. Morning clients are read live and not stored |
| Payment records | 7 years (Israeli tax law requirements) |
| Analytics data | Aggregated and anonymized; retained for up to 24 months |
| Session recordings | Deleted automatically after 90 days |
| MCP connectors | Nothing stored by the connector itself; conversation messages held in memory for up to 10 minutes; operational logs 30 days; usage records may be kept longer, with no fixed deletion period |
| Backups | Encrypted, rolling 30-day window |
We implement appropriate technical and organizational measures to protect personal data, including:
Under the GDPR (where applicable) and Israeli privacy law, you have the following rights:
If you are an end-user of a business that uses AssistantLabs, that business is the controller of your data; please direct your request to it. We will assist our Customer in responding. To exercise rights regarding data for which AssistantLabs is the controller, contact legal@assistantlabs.io. We will respond within 30 days.
These steps are also published at assistantlabs.io/data-deletion.
Billing and invoice records are kept for 7 years to meet Israeli tax law, and deleted data leaves our backups as they roll over. If you are an end-user, ask the business you spoke to; if you cannot reach it, write to us and we will pass your request on.
The Service uses Firebase Analytics to collect usage data such as page views, session duration and feature interactions. Our own dashboard and marketing web pages also use the Meta Pixel to measure the effectiveness of our marketing and, where relevant, to build advertising audiences for our own promotion of the Service; the Meta Pixel operates only on our web pages visited by account users and prospects, and is never applied to end-user conversations processed on a Customer's behalf. Where required by applicable law, non-essential cookies and the Meta Pixel are set only after consent, and you can withdraw consent or disable them via your browser settings.
Advertising measurement. If you accepted the cookie banner that names ad measurement with Meta, and you then sign up to the Service, our servers report the sign-up to Meta Platforms Ireland Ltd. The report contains your email address and account identifier, both hashed (SHA-256), the Meta Pixel browser and click identifiers (_fbp, _fbc), your IP address, your browser's user-agent string, and the page you arrived on. Meta uses it to measure and improve our own advertising. We keep the IP address and user-agent string with your account's sign-up record for this purpose only. If you write to us on WhatsApp after clicking one of our ads, we tell Meta whether the conversation became an enquiry, a meeting or a purchase, and the purchase value. We identify the conversation only by the click identifier Meta itself attached to it, and we never send Meta your phone number or the content of your messages. If you did not accept that banner, no sign-up report is sent about you. To object, contact legal@assistantlabs.io.
Results reported by Customers. A Customer who runs Click-to-WhatsApp ads may switch on "Send results to Meta". When they do, we act on their instructions and send their own Meta account the click identifier Meta attached to a conversation that started from their ad, together with the kind of result the Customer chose to report (an enquiry, a meeting or a purchase) and its value. No message content or phone number is sent. It is off unless the Customer switches it on.
First-days session recordings. During the first days after an account is created, we record how its users move through our dashboard — the screens they open, what they click, and moments where something did not work — so we can find and fix the places where new users get stuck. Text typed into forms is masked (except messages written to our Logos setup assistant), passwords and payment fields are never recorded, and screens showing your customers' conversations and contact details are masked. Recordings are stored on our own Google Cloud infrastructure, are accessible only to AssistantLabs staff, are not shared with third parties, and are deleted automatically after 90 days.
Recordings of visits from our ads. When someone reaches our website from one of our Google ads, we may record how they use the page: where they scroll and click, and where something didn't work. We use this to improve our pages. We don't store an IP address, cookie or click identifier with the recording, text typed into forms isn't recorded, and the chat window isn't recorded. We keep the ad's keyword, campaign and ad group. Recordings are deleted after 90 days. Declining cookies in the banner stops the recording.
Email open and click tracking. Email sent through the Service from a campaign or an email template, including in a sales journey, contains a tiny invisible image and links that pass through our server. When the image loads we record that the email was opened, and when a link is followed we record the click and the link, then send the reader on to the page. WhatsApp campaign links are counted the same way. These records sit against the recipient in the Customer's workspace, are shown to the Customer and are reachable through a connector, and are used to report on the campaign and to decide what that campaign sends next. We do not store an IP address or device details with them. Some mail apps load images automatically, so an open is an estimate. Every such email carries an unsubscribe link. The Customer is the controller of this data and is responsible for telling its recipients about it; when we send our own marketing email this way, we are the controller. Replies an assistant sends on the email channel, and account and service email, are not tracked this way.
The web chat widget may set cookies or local-storage entries for session management. Customers are responsible for obtaining any cookie consent required under applicable law (e.g., the ePrivacy Directive) when embedding the web chat on their websites.
AssistantLabs accounts are for businesses, and the person who accepts our terms for a business must be at least 18. The Service is not directed at children and contains no mature content. The people who message a business through the Service are that business's end-users, and the Customer controls their data. We do not knowingly open an account for a child. If we learn that one was opened, we will close it and delete its data promptly.
We may update this Privacy Policy from time to time. We will notify you of material changes by email at least fourteen (14) days before they take effect. The "Effective Date" at the top indicates the most recent revision.
If you have questions, concerns or requests regarding this Privacy Policy or our data practices, please contact:
AssistantLabs – Legal & Privacy
Email: legal@assistantlabs.io
Address: Ha'Aniya Arinpura 8, Netanya, Israel
If you are not satisfied with our response, you may lodge a complaint with the Israeli Privacy Protection Authority (PPA) or, for EU residents, your local supervisory authority.