Privacy Policy
This Privacy Policy explains how Talari AI Services P.S.A. (“Talari”, “we”, “us”) collects, uses, stores, and protects your personal data when you use the talari.ai platform and related services. It is drafted in accordance with Regulation (EU) 2016/679 (GDPR), Articles 13 and 14.
Version 1.3 replaces version 1.2 of 2 November 2026 and corrects how this policy describes pseudonymisation (mainly sections 5 and 7). On 20 November 2026 we also replaced Flynnt UG (apertus.ai, Germany) with Infomaniak Network SA (Switzerland) for relevance ranking. Version 1.2 replaced version 1.1 of 4 September 2026. What changed, and when: between 1 and 9 October 2026, speech-to-text and the AI language models moved to Scaleway SAS (France). On 30 October 2026 we added OVHcloud SAS as a backup speech-to-text provider, and IONOS SE as a backup provider for AI language models, relevance ranking and search embeddings. On 2 November 2026 we added how long our main AI provider may keep a request to investigate an error or misuse (at most 14 days). The processors listed below are the ones we use today; Berget AI AB is listed but not yet in use. Contrary to what we had promised, we did not notify registered users before these changes. If you object to one of the processors, write to legal@talari.ai; if we cannot resolve your objection, you may close your account free of charge.
1. Identity and Contact Details of the Controller
The controller of your personal data within the meaning of Art. 4(7) GDPR is:
TALARI AI SERVICES PROSTA SPÓŁKA AKCYJNA
KRS: 0001264352
NIP: 9522289458
Registered address: ul. Ignacego Paderewskiego 144A lok. 2, 04-438 Warszawa, Poland
Email: legal@talari.ai
Website: talari.ai
Talari AI Services P.S.A. is a simple joint-stock company (prosta spółka akcyjna, P.S.A.) under Polish law, registered in the Polish National Court Register (KRS) under number 0001264352, kept by the Sąd Rejonowy dla m.st. Warszawy w Warszawie, XIV Wydział Gospodarczy Krajowego Rejestru Sądowego (District Court for the Capital City of Warsaw, 14th Commercial Division of the National Court Register).
2. Data Protection Contact
We have not designated a Data Protection Officer within the meaning of Art. 37 GDPR. For any question about how we process your personal data, or to exercise your rights under the GDPR, contact:
Data protection contact
Email: trust@talari.ai
We respond within one month of receipt, in accordance with Art. 12(3) GDPR. Where a request is complex or numerous that period may be extended by a further two months, and we will tell you within the first month if that happens.
3. Categories of Personal Data
Depending on how you use our platform, we may collect and process the following categories of personal data:
- Account data — name, email address, username, role, tenant affiliation.
- Session transcription data — audio converted to text during NinaSession documentation sessions.
- Pseudonymisation mappings — lookup tables linking personally identifiable information (PII) to placeholder tokens, used to pseudonymise transcripts before AI processing; stored encrypted.
- AI analysis outputs — structured reports generated by AI models from session transcripts (pseudonymised by default).
- Session metadata — timestamps, session duration, session mode (solo, remote, text), language settings.
- Remote session guest data — consent records, microphone permissions, IP addresses, and WebRTC connection metadata for guests joining via invite link.
- Uploaded documents — professional context documents (e.g. client histories, reference materials) uploaded for retrieval-augmented generation (RAG).
- AI chat and knowledge base — the content of your conversations with the AI assistant (chat history) and the documents you upload to your libraries (knowledge base).
- Template filler — document templates, drafts and filled documents.
- Chat widgets on customer websites — conversation statistics (number of messages, language, site domain) and contact details entered in the widget form (name, email, phone).
- Consent records — your consent decisions (e.g. in the cookie banner), with a hash of your IP address and a shortened browser identifier.
- Billing data — subscription tier, payment method identifiers, invoicing data. Full payment card details are processed exclusively by Stripe and never stored on our servers.
- Authentication data — hashed passwords (bcrypt), JWT tokens, Redis session identifiers, two-factor authentication secrets.
- Audit logs — records of login events, administrative actions, data access, and system events for security monitoring.
- LLM provider interaction data — prompts and responses exchanged with third-party AI model providers (section 5 explains which text is pseudonymised).
4. Purposes of Processing and Legal Basis
The following table sets out the specific purposes for which we process your personal data, the legal basis under the GDPR, and the applicable retention period:
| Data | Purpose | Legal Basis | Retention |
|---|---|---|---|
| Account data (name, email, role) | User registration and authentication | Art. 6(1)(b) GDPR | Duration of account |
| Session transcription (audio-to-text) | Real-time documentation of professional sessions | Art. 6(1)(b); Art. 9(2)(h) if health data | Until the session is finalised or deleted |
| Pseudonymisation mapping | Replace identifiable data with tokens before LLM processing | Art. 6(1)(c) GDPR | Until the session is finalised or deleted (stored encrypted) |
| AI analysis and reports | Automated report generation from session transcripts (pseudonymised by default) | Art. 6(1)(b) GDPR | Until the session is finalised or deleted |
| Session metadata (timestamps, duration, mode) | Platform operation, usage tracking, billing | Art. 6(1)(b) GDPR | Retained after finalisation |
| Remote session guest data (consent, microphone, IP) | Enable remote participation via WebRTC | Art. 6(1)(a) GDPR (consent) | Until the session is finalised or deleted; guest display name until the session is deleted |
| Uploaded context documents | Contextual retrieval for session analysis | Art. 6(1)(b) GDPR | Until the document is deleted or the session is finalised or deleted |
| AI chat (conversation content) | AI assistant answers and conversation history | Art. 6(1)(b) GDPR | 180 days after the last activity in the conversation (organisation setting: 30–365 days) |
| Library documents (knowledge base) | AI answers based on your own documents | Art. 6(1)(b) GDPR | Until the document or the account is deleted |
| Template filler (drafts, filled documents) | AI-assisted completion of document templates | Art. 6(1)(b) GDPR | Drafts: 30 days after the last save; filled documents: up to 30 days (chosen by the user) |
| Billing and payment data | Subscription management and invoicing via Stripe | Art. 6(1)(b) GDPR + PCI-DSS | As required by tax law |
| Authentication tokens (JWT, Redis sessions) | Secure session management | Art. 6(1)(c) GDPR | Up to 8 hours (login session) |
| Audit logs (login, actions, IP) | Security monitoring and compliance | Art. 6(1)(c) GDPR | 24 months |
| LLM provider calls (prompts; see section 5 on pseudonymisation) | AI report generation and analysis | Art. 6(1)(b) GDPR | Not stored by the provider. Exception: our main AI provider (Scaleway) may keep a request for up to 14 days to investigate an error or misuse |
Where we rely on legitimate interest as a legal basis, we have conducted a balancing test to ensure that our interests do not override your fundamental rights and freedoms.
5. Special Category Data (Article 9)
Health and Sensitive Data
When talari.ai is used in healthcare, therapeutic, or counselling contexts, session transcriptions may contain special categories of personal data within the meaning of Art. 9(1) GDPR, including data concerning health.
We process such data under Art. 9(2)(h) GDPR, which permits processing necessary for the purposes of preventive or occupational medicine, medical diagnosis, the provision of health or social care, or the management of health or social care systems and services. In such cases, the professional user of talari.ai acts as the Data Controller for their clients' data, and Talari acts as the Data Processor.
By default, the session transcript is pseudonymised (identifiers replaced with placeholders) before an AI language model writes a report from it; the same applies to earlier client documents used to pre-fill a template and to the client context in AI chat. Automatic detection can miss identifiers. The following are processed as entered, without pseudonymisation: session audio, which is sent to our EU speech-to-text provider for transcription only; messages and attachments in AI chat; documents attached in the Template Filler; the live Virtual Assistant interview; and passages from your own uploaded documents sent for relevance ranking. In our standard configuration, they are processed by the processors listed in section 7, which are located in the EU or, for relevance ranking, in Switzerland. The pseudonymisation mapping is maintained solely within the user's tenant environment and is permanently purged when the session is finalised.
6. Automated Processing and Profiling
talari.ai employs automated processing, including AI-powered analysis, to generate reports and documentation from session transcripts. This processing involves:
- Automatic transcription of audio to text using speech-to-text models.
- Automatic detection and pseudonymisation of identifiers (section 5 explains what is covered).
- AI-generated structured reports based on session transcripts (pseudonymised by default) and professional templates.
This automated processing does not constitute profiling within the meaning of Art. 22 GDPR. No decisions with legal or similarly significant effects are made solely on the basis of automated processing. All AI-generated outputs are presented as drafts for professional review and are not intended to replace clinical, legal, or other professional judgement.
7. Third-Party Processors
We engage the following processors to deliver our services:
| Processor | Country | Purpose | Safeguard |
|---|---|---|---|
| Hetzner Online GmbH | Germany | Server hosting and infrastructure (all platform data) | Processing in the EU |
| Scaleway SAS | France | AI language models (text; see section 5 on pseudonymisation) and speech-to-text (session audio) | Processing in the EU; data processing agreement (Art. 28 GDPR) |
| OVHcloud SAS | France | Backup speech-to-text (session audio), used only when Scaleway is unavailable | Processing in the EU; data processing agreement (Art. 28 GDPR) |
| IONOS SE | Germany | Backup AI language models (text; see section 5 on pseudonymisation), relevance ranking (search query and retrieved passages) and search embeddings (search query text), used only when the primary service is unavailable | Processing in the EU; data processing agreement (Art. 28 GDPR) |
| Infomaniak Network SA | Switzerland | Relevance ranking of retrieved document passages for AI answers (search query and passages) | EU adequacy decision for Switzerland (Art. 45 GDPR); data processing agreement (Art. 28 GDPR) |
| Berget AI AB | Sweden | AI models and relevance ranking (second EU provider, being introduced; not yet in use) | Processing in the EEA |
| Stripe Inc. | USA | Payment processing and subscription billing | SCCs + adequacy decision (DPF) |
| Namecheap Inc. (Private Email) | USA | Transactional email delivery, including AI-chat transcripts you have sent to your own address | SCCs per Art. 46 GDPR |
| Google (reCAPTCHA) | Ireland / USA | Bot protection on sign-up and application forms | Adequacy decision (DPF), SCCs |
| Telegram Bot API | United Arab Emirates | Messaging integration (optional) | Only if enabled by the user |
Speech-to-text and AI language models are provided by processors in the European Union. Relevance ranking is provided by Infomaniak Network SA in Switzerland, for which the European Commission has adopted an adequacy decision. Session transcripts are pseudonymised by default before a language model writes a report from them (section 5 lists what is not pseudonymised); audio is sent to our speech-to-text provider for transcription only. Processors in the United States receive account, billing and form-security data and the content of the e-mails we send you, including an AI-chat transcript if you have it sent to your own e-mail address. Our Trust page lists the processors in use and shows where a data processing agreement (Art. 28 GDPR) is still to be concluded.
8. International Data Transfers
Certain sub-processors are located outside the European Economic Area (EEA), specifically in Switzerland (relevance ranking), the United States and, for the optional Telegram integration, the United Arab Emirates. For these transfers, we rely on the following safeguards pursuant to Chapter V of the GDPR:
- Standard Contractual Clauses (SCCs) adopted by the European Commission under Art. 46(2)(c) GDPR, agreed with US-based processors where the Data Privacy Framework does not apply.
- EU-U.S. Data Privacy Framework (DPF) where the processor is certified under the adequacy decision of 10 July 2023 (applicable to Stripe and Google).
- Adequacy decision for Switzerland— Commission Decision 2000/518/EC (Art. 45 GDPR), for Infomaniak Network SA (relevance ranking).
- AI processing in the EU and Switzerland— speech-to-text and AI language models are provided by processors in the European Union, relevance ranking by Infomaniak in Switzerland (see Section 7). US-based processors receive the data described in Section 7.
Copies of the applicable SCCs and transfer impact assessments are available upon request from trust@talari.ai.
9. Data Retention
We retain personal data only as long as necessary for the purposes set out in this policy:
- Active and saved sessions: Session data (transcripts, pseudonymisation mappings, AI analysis, reports, context documents) is retained until you finalise or delete the session, or delete your account. We do not delete it automatically after a fixed period.
- Finalised sessions: Upon finalisation, all session content (transcripts, context documents, reports, PII mappings) is permanently and irreversibly purged. Only metadata (timestamps, duration, mode, a remote guest's display name) is retained.
- Audio: Recordings are not stored. Audio is used for transcription only and discarded immediately afterwards.
- Context documents: Retained until you delete the document, or finalise or delete the session.
- AI chat history:180 days after the last activity in a conversation; an organisation's administrator can set a period between 30 and 365 days.
- Library documents (knowledge base): Until you delete the document or your account.
- Template filler: Drafts are kept for 30 days after the last save; filled documents for the period chosen when the document is closed, at most 30 days.
- Login sessions: A login session lasts up to 8 hours.
- Audit logs: Security and administrative audit logs are retained for 24 months.
- Consent records: For as long as the consent applies and for 7 years after it is withdrawn, so that we can demonstrate consent.
- Backups: Encrypted backups of the database and files are kept for up to 30 days. Data deleted from the platform disappears from the backups after that time at the latest.
- Account data: Retained for the duration of your account. Upon account deletion, personal data is purged within 30 days, subject to legal retention obligations.
10. Your Rights Under the GDPR (Articles 15–22)
You have the following rights regarding your personal data:
- Right of access (Art. 15)— obtain confirmation of whether we process your personal data and receive a copy of that data.
- Right to rectification (Art. 16)— request correction of inaccurate personal data.
- Right to erasure (Art. 17)— request deletion of your personal data. Note that session finalisation constitutes an irreversible erasure mechanism built into the platform.
- Right to restriction of processing (Art. 18)— request limitation of processing in certain circumstances.
- Right to data portability (Art. 20)— receive your personal data in a structured, commonly used, and machine-readable format.
- Right to object (Art. 21)— object to processing based on legitimate interest.
- Right to withdraw consent (Art. 7(3))— where processing is based on consent (e.g. remote session guest participation), you may withdraw consent at any time without affecting the lawfulness of processing carried out prior to withdrawal.
- Right not to be subject to automated decision-making (Art. 22)— our AI processing does not produce decisions with legal or similarly significant effects. All outputs are professional drafts subject to human review.
To exercise any of these rights, please contact us at trust@talari.ai. We will respond within 30 calendar days of receipt of your request, in accordance with Art. 12(3) GDPR. This period may be extended by a further 60 days for complex or numerous requests, in which case we will inform you within the initial 30-day period.
11. Right to Lodge a Complaint
If you believe that our processing of your personal data infringes the GDPR, you have the right to lodge a complaint with a supervisory authority pursuant to Art. 77 GDPR.
As our company is established in Poland, the lead supervisory authority is:
Urząd Ochrony Danych Osobowych (UODO)
Polish Data Protection Authority
Website: uodo.gov.pl
Email: kancelaria@uodo.gov.pl
You also have the right to lodge a complaint with the supervisory authority of the EU/EEA Member State of your habitual residence or place of work.
13. Changes to This Privacy Policy
We may update this Privacy Policy from time to time to reflect changes in our processing activities, legal requirements, or platform features. Material changes will be communicated via email notification to registered users and prominently displayed on the platform at least 14 days before they take effect.
The “Last updated” date at the top of this document indicates the date of the most recent revision. We encourage you to review this policy periodically.