1. Who we are and how to contact us
Data Controller
TMMA SRLS, VAT IT17152861005, Via Durazzo 28, 00195 Roma (RM), Italy. General inquiries: [email protected]. Privacy matters: [email protected].
Data Processor
When MessageMind processes End User personal data on behalf of a Customer, TMMA SRLS acts as Data Processor strictly on the Customer's documented instructions, under a written Data Processing Agreement (DPA) that implements Article 28 GDPR.
EU location & supervisory authority
The Controller is established in Italy. The competent supervisory authority is the Garante per la protezione dei dati personali (GPDP), Piazza Venezia 11, 00187 Roma, Italy.
Data Protection Officer
A DPO has not been formally appointed. The need for a DPO is reassessed routinely as the platform evolves and as processing activities expand.
2. Scope & relationship with other documents
This Privacy Policy complements the MessageMind Terms & Conditions and the Data Processing Agreement (DPA) signed with each Customer. It covers the MessageMind public websites, the platform, the APIs, the chatbot runtime, and every integration we operate.
Cookies and similar technologies are governed by a separate Cookie Policy. Sub-processors are listed in a Sub-processor List that is maintained and updated as the vendor stack changes.
3. Key definitions
Customer: the business that subscribes to MessageMind and configures the platform for its own use.
End User: the individual that interacts with a Customer through MessageMind (e.g. a guest, lead, patient, buyer).
Personal Data: any information relating to an identified or identifiable natural person, as defined by Article 4 GDPR.
Special Categories: data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, trade union membership, genetic data, biometric data, data concerning health, sex life or sexual orientation.
Processing: any operation performed on Personal Data, automated or not (collection, storage, retrieval, transfer, erasure).
4. What we collect
A. Data you provide directly
Account information (name, email, phone, role), billing information (company name, VAT, billing address, payment instrument metadata), support correspondence (messages, attachments, tickets), and explicit consents you give for marketing, recording or AI features.
B. Data processed through the Platform
Conversations exchanged between End Users and the Customer (messages, transcripts, attachments, status), integration metadata (channel identifiers, OAuth tokens, webhook payloads), and the chatbot/AI inputs and outputs needed to deliver replies.
C. Data collected automatically
Technical logs (IP, user agent, timestamps, request and response metadata) and product analytics (pseudonymous events used to understand usage and improve reliability).
D. Special categories
We do not seek special categories of data. If an End User voluntarily shares them in a conversation, we apply additional safeguards (access restriction, encryption at rest, minimum retention) and rely on the appropriate Article 9 exception.
E. Children
The Services are not directed to children. The minimum age in Italy is 14 years for direct consent. Below that age, processing relies on parental authorization. Customers are responsible for age-gating where their own audience requires it.
5. Why we process data (purposes) and legal bases
We rely on the legal bases of Article 6 GDPR: performance of a contract Art. 6(1)(b), compliance with a legal obligation Art. 6(1)(c), legitimate interests Art. 6(1)(f), and consent Art. 6(1)(a) where consent is the appropriate basis (marketing, non-essential cookies, recording).
Special categories are processed only where an Article 9 exception applies, for example explicit consent or the establishment, exercise or defence of legal claims.
6. Meeting, call, and video recordings
MessageMind never records calls, meetings or video by default. Recording is an opt-in feature and runs only after an informed, explicit consent has been captured from every participant.
Purposes
Quality assurance, training, dispute resolution, transcription and, when explicitly enabled by the Customer, downstream AI processing such as summarization and CRM enrichment.
Controls
A visible banner or audio notice is shown before recording starts. Participants can refuse, and the system must offer a non-recorded path where this is technically feasible.
Storage & access
Recordings are stored encrypted at rest, with role-based access limited to authorized Customer users and to MessageMind staff that strictly need them to deliver support.
Third-party meeting tools
When recordings are produced through third-party meeting tools (e.g. video conferencing platforms), the tool's own privacy notice applies in addition to this Policy.
7. What we do with data
- Provide, operate and maintain the Services that the Customer has subscribed to.
- Route, store and deliver conversations between End Users and Customers across channels.
- Run AI features the Customer has enabled (replies, summaries, qualification, scheduling).
- Maintain security, prevent fraud and abuse, debug failures and protect availability.
- Comply with legal obligations (tax, accounting, lawful requests, records of processing).
- Improve reliability, quality and product performance using aggregated or pseudonymous signals.
- Communicate with the Customer about the Services, including service and security notices.
We do not sell Personal Data. We do not use End User content to train third-party AI models by default. Any exception requires the Customer's explicit, documented opt-in.
8. Sharing of data
8.1 Processors and sub-processors
We rely on infrastructure, communication and AI vendors that act as our sub-processors under written agreements. The current list is published in the Sub-processor List and updated as the stack changes.
8.2 Customer-enabled channel partners & integrations
When the Customer connects a third-party channel or system, data flows to and from that provider under its own privacy notice. This includes Meta (Facebook, Instagram, WhatsApp Business), Twilio, Gmail, Outlook/Microsoft 365, Shopify, WooCommerce, Squarespace, Calendly, Stripe, Revolut, Zapier, Make and n8n, among others.
8.3 Limited-use commitments for certain APIs
For APIs such as Gmail and Microsoft Graph we honor the provider's limited-use requirements: data accessed through those APIs is used only to deliver the user-facing feature, is not transferred to others except as needed for the feature, is not used for advertising, and is not read by humans except for security, legal or with the user's explicit consent.
8.4 Professional advisers & authorities
We may disclose data to auditors, lawyers, accountants, insurers and competent authorities when this is necessary to defend rights, comply with the law, or respond to a binding request.
8.5 Corporate transactions
In the context of a merger, acquisition or asset transfer, data may be shared with the counterparty under confidentiality. We will notify Customers and, where required, End Users.
9. Marketing, soft-spam, telemarketing & RPO
Email, SMS and WhatsApp marketing is sent only on a valid consent. Every message offers a simple, free opt-out, and we honor it promptly.
Soft-spam (limited promotional communications about similar products to existing customers) is permitted under art. 130(4) Codice Privacy, with the same opt-out at any time and at the point of collection.
Telemarketing campaigns respect the Italian Registro Pubblico delle Opposizioni (RPO). Numbers registered in the RPO are filtered out before any campaign is launched.
10. Cookies, SDKs & similar technologies
Strictly necessary cookies are set without consent because they are required to deliver the Services. Non-essential cookies (analytics, marketing, profiling) are activated only after the user grants consent via the cookie banner.
We do not use dark patterns. The cookie banner offers a clear "accept", "reject" and "preferences" choice with equivalent prominence. Closing the banner does not equal consent, and all non-essential technologies remain off.
We do not perform device fingerprinting without consent. Full details, including names, durations and providers, are in the separate Cookie Policy.
11. AI features & model providers
11.1 Scope of AI processing
The platform uses large language and audio models to generate replies, transcribe audio, summarize conversations, classify intent and assist agents. AI is invoked only on the features the Customer has enabled.
11.2 No training / no sharing pledge
By default, End User content is not used to train MessageMind or third-party models. Sub-processors are contracted to disable training on our prompts and completions where the option exists. Any exception requires the Customer's explicit, written opt-in.
11.3 Data minimization & redaction
Where feasible we redact direct identifiers (emails, phone numbers, card numbers) from prompts that are not strictly required to answer. Prompt size and context are kept to the minimum needed.
11.4 Calls, audio & recordings
Audio is processed only when the Customer enables voice features. Recordings, transcripts and derived embeddings inherit the retention rules of the corresponding feature and are deleted on request, subject to legal-hold exceptions.
11.5 Custom voice cloning (optional)
Lawful basis & consent. Voice cloning is offered only on the explicit consent of the voice owner, with a clear description of purpose, duration and revocation rights.
Purpose limitation. A cloned voice is used solely for the purposes the voice owner has agreed to, on the specific MessageMind tenant that requested the clone.
Isolation. Voice models are tenant-isolated and not shared across Customers. They are not used to fine-tune third-party foundation models.
No impersonation / no identification. Cloned voices may not be used to impersonate other individuals, to bypass authentication or to circumvent voice-based identification.
Inputs & artifacts. Source audio, prompts, transcripts and intermediate artifacts are stored encrypted, with access strictly limited to engineers that need them for delivery or support.
Retention & deletion. Voice clones are deleted when consent is revoked, when the underlying contract ends, or at the end of the retention window agreed with the voice owner.
Transparency to End Users. When a cloned voice speaks to End Users, the Customer is responsible for the disclosure required by applicable law (e.g. AI Act transparency obligations).
11.6 Vendor roles & contracts
AI vendors act as our sub-processors under signed DPAs that implement Article 28 GDPR. The vendor list and roles are in the Sub-processor List.
11.7 Human oversight, fairness & safety
Customer admins can review, override or disable AI outputs. Sensitive flows (e.g. medical, financial) include guardrails and escalation paths to humans. Outputs are evaluated periodically for safety, accuracy and bias.
12. International data transfers
We prefer EU/EEA primary storage. When a transfer outside the EEA is needed (e.g. to a model provider in the US), we rely on the Standard Contractual Clauses (SCCs), the EU-US Data Privacy Framework where the recipient is certified, and on the supplementary measures appropriate to the destination.
13. Security
We protect the platform with TLS 1.2+ in transit, AES-256 encryption at rest, role-based access control, secure SDLC practices (code review, dependency scanning, secret management), business continuity procedures and periodic staff training.
In the event of a personal data breach, we notify the competent supervisory authority within 72 hours where the breach is likely to result in a risk to individuals, and we inform affected data subjects without undue delay where the risk is high.
14. Retention periods
We keep data only as long as it is necessary for the purposes described, plus the minimum period required by law. The table below summarizes our defaults; Customer-configurable items can be overridden in the relevant settings.
| Data category | Typical retention | Notes |
|---|---|---|
| Customer account data | Contract term + 24 months | Deleted sooner on verified request, unless required for legal claims. |
| OAuth tokens & integration configs | While integration active + 30 days | Tokens encrypted; revocation deletes access immediately. |
| API / webhook call logs | 12 to 18 months | For reliability, security and audit. |
| Billing & invoices | 10 years | Italian civil and fiscal law. |
| Support tickets & emails | 36 months | May be extended for legal claims. |
| Meeting / call recordings | 12 months | Unless earlier deletion is requested. |
| Conversation logs (End Users) | Customer-configurable; default 24 months | The Controller may override. |
| E-commerce data (Shopify / Woo / Squarespace) | Customer-configurable; default 24 months | Excludes full card data. |
| Scheduling data (Calendly) | Customer-configurable; default 24 months | Invitee name, email, time slots. |
| Payment event metadata (Stripe / Revolut) | Customer-configurable; typical 36 months | No full PAN / CVV. |
| Security logs | 12 months | Extended if an investigation is ongoing. |
| Product analytics (pseudonymous) | 24 months | Shorter if consent is withdrawn. |
| Cookie identifiers | Per Cookie Policy | Respect device settings. |
15. Your rights
Subject to the conditions of Articles 15 to 22 GDPR, you have the right to access your data, to rectify it, to erase it, to restrict its processing, to data portability, to object to processing based on legitimate interests, to withdraw consent where consent is the legal basis, and not to be subject to a decision based solely on automated processing that produces legal or similarly significant effects on you.
How to exercise these rights
Write to [email protected]. End Users should contact the Customer first; we will route the request to the Customer and assist as Processor.
Complaints
You may lodge a complaint with the Garante per la protezione dei dati personali, or with the supervisory authority of your habitual residence, place of work or place of the alleged infringement.
16. Controller and Processor responsibilities
When MessageMind is a Controller (websites, marketing, billing, internal operations), it determines purposes and means of processing. When MessageMind is a Processor (delivery of the Services to a Customer), the Customer determines purposes and means within the boundaries of the DPA and the platform's configuration.
Each party is responsible for its share of obligations: notices to End Users, lawful collection, audience configuration, channel-specific rules, retention overrides, and the response to data subject requests.
17. Integrations & channel-specific disclosures
17.1 E-commerce platforms
Shopify, WooCommerce and Squarespace Commerce: order, customer and conversation metadata flows to enable order-aware replies, status updates and post-purchase support. Card data is never fetched; payment processors handle that directly.
17.2 Scheduling
Calendly: invitee name, email, time slot and the answers to the booking form are processed to confirm meetings and to enrich the conversation context.
17.3 Payments
Stripe and Revolut: we process payment event metadata (status, identifiers, amounts) to confirm transactions and to power receipts. We do not store full card numbers or CVV.
17.4 Email & webmail connectors
Custom SMTP/IMAP, Gmail and Outlook/Microsoft 365: the platform accesses inbox messages within the scope authorized at OAuth time, to deliver inbox-aware features. Limited-use rules apply where the provider requires them.
17.5 Telecoms & messaging
Twilio for SMS and voice, WhatsApp via Meta or Twilio for messaging: phone numbers, message content and call metadata are processed to deliver the conversation across these channels.
17.6 Automation hubs & custom APIs
Zapier, Make / Integromat, n8n and custom webhooks: the data shared depends on the workflow the Customer configures. The Customer is responsible for the lawfulness of the data fields that travel through these hubs.
17.7 Other knowledge & storage connectors
Knowledge bases, file storage and CRM connectors are processed within the scope authorized at connect time, and only to deliver the feature the Customer has activated.
18. Automated decision-making & profiling
The platform does not make decisions producing legal or similarly significant effects without human involvement. AI-suggested replies and lead scoring are advisory; Customer admins can review, override or disable them.
19. Data protection by design & DPIAs
Privacy by design and by default is embedded in our development lifecycle. Where a new feature is likely to result in a high risk to the rights and freedoms of individuals, we run a Data Protection Impact Assessment (DPIA) and consult the supervisory authority where required.
20. International users
If you access the Services from outside the EU/EEA, you acknowledge that data may be processed in the EU/EEA and, where necessary, in jurisdictions outside the EEA under the safeguards described in Section 12.
21. Changes to this Policy
We update this Policy from time to time. The version date at the top of the page reflects the latest update. Material changes are communicated in advance through the platform or by email.
22. Contact
Privacy matters: [email protected].
TMMA SRLS, Via Durazzo 28, 00195 Roma (RM), Italy.
Garante per la protezione dei dati personali: Piazza Venezia 11, 00187 Roma, Italy.
23. Governance, accountability & audits
We maintain internal policies on information security, data protection, vendor management, incident response, retention and acceptable use. Controls are reviewed periodically and adapted to the evolving threat landscape and to the scope of the platform.
Customers with a contractual right to audit may exercise it under the conditions agreed in the DPA, with reasonable notice and under confidentiality.
24. Data classification & handling rules
Data is classified as public, internal, confidential or restricted. Handling rules (storage, transit, sharing, retention, disposal) are aligned to the classification. Restricted data (e.g. credentials, secrets, special categories) receives the strongest protections.
25. Technical & organizational security measures (extended)
Network segmentation, principle of least privilege, secret management with rotation, dependency vulnerability scanning, hardened CI/CD, code review, audit logging, backup with point-in-time recovery, periodic restore tests and tabletop exercises for incident response.
Workstation security: full-disk encryption, MDM enrollment, endpoint detection, password manager, hardware-backed multi-factor authentication on critical systems.
26. Government & law-enforcement requests
We respond only to requests that are valid, lawful and properly addressed. We push back on requests that are overbroad or that do not meet the legal requirements of the jurisdiction. Where lawful, we notify the affected Customer before disclosure.
27. Data subject request (DSR) handling
We acknowledge DSRs without undue delay and reply within one month, extendable by two months where the request is complex or numerous. We verify the requester's identity in a proportionate way, route the request to the Customer when MessageMind is Processor, and keep an internal log to demonstrate accountability.
28. Consent management, cookies & preference center (extended)
We use a consent management platform that records, on a per-user basis, the categories accepted or refused, the timestamp, the text version of the banner and the way consent was collected. Users can change their preferences at any time through the preference center.
29. Advanced integration details & scopes (deep dive)
29.1 Shopify
OAuth scopes are limited to what the activated features require. Typical reads include orders, customers and conversation metadata. Writes (e.g. tagging an order, posting a note) are guarded by the Customer's configuration.
29.2 WooCommerce
Connected via REST API keys with the minimum permissions needed. Order and customer reads enable order-aware replies and proactive messages.
29.3 Squarespace
Connected via the Squarespace Commerce APIs to read orders and customer metadata for support and post-purchase workflows.
29.4 Calendly
OAuth scopes cover invitees, scheduled events and webhooks for booking lifecycle. Used to read invitee data and to power scheduling-aware replies.
29.5 Stripe
Restricted API key or OAuth with read access to payment events and customers. Used to confirm payments and to provide receipt context inside conversations.
29.6 Revolut
Business API access scoped to payment event reads. Same purpose as Stripe: confirmations and receipts in-conversation.
29.7 Gmail / Outlook
OAuth scopes are minimized to what the connector needs (read, send, modify labels). Limited-use rules apply: data is not used for advertising, is not transferred outside the feature, and is not human-read except for security, legal or explicit consent.
29.8 Twilio & WhatsApp
Phone numbers, SMS payloads, voice call metadata and recordings (when enabled) flow through Twilio. WhatsApp business messaging is governed by Meta's policies and by template approvals.
29.9 Zapier / Make / n8n
Webhooks and triggers exchange the fields the Customer maps in their workflow. The Customer is responsible for the lawfulness of the data sent or received through these automations.
30. AI transparency, fairness & evaluation
We document the intended purpose and limits of each AI feature. Outputs are evaluated periodically against quality, safety and bias criteria; failure modes drive guardrails and retraining of prompts or routing logic. End User-facing AI is disclosed where the law requires it.
31. Children & parental authorization (expanded)
The platform is not designed for children under the Italian minimum age (14). When a Customer's audience includes minors, the Customer is responsible for the age-gating, the parental authorization mechanism and the related notices. We provide configurable safeguards to support this.
32. Whistleblowing & reporting channels
Concerns about the lawfulness of processing, security incidents or breach of this Policy can be reported confidentially to [email protected]. We protect reporters from retaliation in line with applicable whistleblowing law.
33. Data portability & export formats
Customers can export account, conversation and configuration data in structured, machine-readable formats (typically JSON or CSV). End User portability requests are handled by the Customer with our assistance as Processor.
34. Service-specific terms & precedence
Some features and integrations are governed by additional terms (e.g. a feature-specific DPA addendum, a vendor's own terms). In the event of conflict, the more specific document prevails for that particular feature, while this Policy remains the general framework.
Annex A. Examples of processing activities (Records of Processing overview)
| Activity | Categories of data | Recipients | Retention |
|---|---|---|---|
| Account management | Identity, contact, billing | Internal teams, accountant | Contract + 24 months |
| Conversation delivery | Messages, attachments, identifiers | Channel partners, infra vendors | Customer-configurable, default 24 months |
| AI replies & summaries | Prompts, completions, transcripts | Model providers (sub-processors) | As per feature, no training by default |
| Billing & invoicing | Company, VAT, invoice content | Payment processor, tax authorities | 10 years (fiscal law) |
| Support | Ticket content, attachments | Support tooling, internal | 36 months |
| Security monitoring | IP, user-agent, audit logs | SIEM vendor, internal | 12 months |
Annex J. Lawful-basis matrix (examples)
| Processing | Legal basis | Notes |
|---|---|---|
| Account creation & service delivery | Contract — Art. 6(1)(b) | Required to provide the Services. |
| Billing & invoicing | Legal obligation — Art. 6(1)(c) | Italian civil and fiscal law. |
| Security & fraud prevention | Legitimate interests — Art. 6(1)(f) | Balanced against user rights. |
| Product analytics (pseudonymous) | Legitimate interests — Art. 6(1)(f) | Where consent is not required. |
| Marketing emails & SMS | Consent — Art. 6(1)(a) | Free, granular, withdrawable. |
| Soft-spam to existing customers | Art. 130(4) Codice Privacy | Opt-out at any time. |
| Non-essential cookies | Consent | Captured via the CMP. |
| Call / meeting recording | Explicit consent | Banner shown before recording. |
| Voice cloning | Explicit consent of the voice owner | Purpose-limited, tenant-isolated. |
| Special categories voluntarily shared | Art. 9 exception (typically explicit consent) | Extra safeguards apply. |
Annex K. Legitimate Interests Assessment (LIA) template (summary)
For each processing relying on legitimate interests we run a three-step test: (1) Purpose, the interest pursued; (2) Necessity, whether processing is necessary and proportionate to the purpose; (3) Balancing, the interests, rights and reasonable expectations of data subjects. The outcome is documented and periodically reviewed.
Annex L. DPIA triggers & checklist
Triggers include: large-scale processing of special categories, systematic monitoring, novel use of AI or biometric data, combining datasets in ways individuals would not reasonably expect, processing involving children at scale, and any feature flagged by a supervisory authority as high-risk.
The DPIA checklist covers: description of processing, necessity and proportionality, risk assessment, controls, residual risk, consultation with the supervisory authority where required.
Annex M. Security control catalog (non-exhaustive)
Identity and access (SSO, MFA, RBAC), encryption (TLS 1.2+, AES-256 at rest), key management with rotation, network segmentation, secure SDLC (code review, SAST, dependency scans), audit logging, backup and disaster recovery, vendor risk management, secure remote work, periodic awareness training, incident response runbooks.
Annex N. Cookie categories & CMP mapping
| Category | Purpose | Consent required | CMP toggle |
|---|---|---|---|
| Strictly necessary | Authentication, security, load balancing, language | No | Always on |
| Functional | Remember preferences, improve experience | Yes | Functional |
| Analytics | Pseudonymous product analytics | Yes | Analytics |
| Marketing | Advertising, attribution, retargeting | Yes | Marketing |
| Profiling | Behavioural profiling for personalization | Yes | Profiling |
Annex O. Article 30 ROPA example record (abridged)
Controller: TMMA SRLS. Activity: AI-assisted reply generation. Categories of data: conversation content, identifiers. Categories of data subjects: End Users of MessageMind Customers. Recipients: hosting and AI sub-processors. Transfers: EU/EEA primary, with SCCs and DPF where applicable. Retention: as per Section 14. Security measures: as per Section 13 and Annex M.
Annex P. Law-enforcement request transparency (policy)
We log the number, type and outcome of law-enforcement requests in aggregate, and we are committed to publishing transparency metrics when the volume becomes material. We notify Customers before disclosure unless prohibited by law.
Annex Q. Model providers & data-handling (placeholders)
The current list of AI sub-processors, the regions in which they process data and the contractual commitments on training and retention is maintained in the Sub-processor List, kept up to date as the stack evolves.
Annex R. Data deletion schedules & backup windows
Logical deletion is propagated to primary stores promptly. Encrypted backups expire on a rolling window (typically 30 days) after which data is unrecoverable. Restoration from backup re-applies the latest deletion log to avoid resurrecting deleted records.
Annex S. AI audio/voice cloning policy & consent language (templates)
S.1 Recording / transcription consent (call banner)
"This call is being recorded and transcribed for quality, training and AI-assisted support. You can object now and we will switch to a non-recorded path where available. For details on how we handle your data see our Privacy Policy."
S.2 Custom voice cloning — explicit consent
"I authorize TMMA SRLS (MessageMind) to create a synthetic voice model from my voice recordings, to be used exclusively by the tenant indicated above, only for the purposes described and within the retention period agreed. I have been informed about my right to withdraw this consent at any time without giving a reason."
S.3 Revocation language
"I revoke the consent previously given to create and use a synthetic voice model from my voice. I understand the voice model will be deleted within the timeframe stated in the Privacy Policy and that any prior, lawful use remains unaffected."
S.4 Prohibited uses
Synthetic voices may not be used to impersonate other individuals, to mislead End Users about the nature of the speaker where the law requires disclosure, to bypass authentication, or for any illegal, discriminatory or harassing purpose.
S.5 Security & access
Voice models, source audio and derived artifacts are encrypted at rest. Access is limited to engineers strictly involved in delivery or support. Access is logged and reviewed periodically.
Questions on this Policy? Write to [email protected].
Back to top