Frequently asked questions

1. Product & getting started

Zentre gives your team one place to use leading AI models — with shared knowledge, agents, and full control over who uses which AI for what. Your platform data is hosted in the EU.

Companies in the EU/DACH region that want to use AI in an organized way — from skilled trades and mid-sized businesses to confidentiality-sensitive sectors such as law firms, tax advisors, healthcare, and financial institutions.

Yes. The service is directed exclusively at businesses within the meaning of Sec. 14 German Civil Code — not consumers.

It currently runs 7 days (up to 5 users, €5 AI credits, 250 MB knowledge storage) and ends automatically — no credit card, no automatic conversion into a paid contract. Without an upgrade, trial data is deleted 30 days after the trial ends; you can export until then. Please don’t process professional secrets (e.g., client or patient data) during the trial — use internal or anonymized data.

Super Admin (full workspace control), Admin (users, teams, models, agents, knowledge bases), Billing Admin (billing and invoices), Member (chat, agents, assigned knowledge bases).

Current models from Mistral, OpenAI, Anthropic (Claude), and Google (Gemini) — delivered through European infrastructure. The current list is at zentre.ai/models. As admin, you decide which models your team may use.

Yes — without losing the conversation context.

PDFs, images, Word documents, Markdown, text files, and other common business formats — up to 20 files per chat, 25 MB each.

Yes, document creation and editing is supported (currently in beta).

Yes. Image generation follows the same rules as the rest of the service: processing via EU endpoints, zero data retention, no training on your content — and in Professional Secrecy Mode the same model classes. Generated image files are marked machine-readably as AI-generated (IPTC Digital Source Type; additionally C2PA Content Credentials where the format supports this). The available image models are shown in the model overview inside the application.

Write to support@zentre.ai. We expand integrations continuously based on customer demand.

One contract, one place, EU hosting — instead of scattered individual accounts with no oversight. Your team uses the right AI per task, shares knowledge and prompts, and you keep control: who uses what, with which data, at what cost.

EZTO TECHNOLOGIES GmbH, Am Brand 41, 55116 Mainz, Germany (AG Mainz HRB 54349).

2. Pricing & Credits

One plan, per user (seat): €21.90 monthly or €18.25 with annual billing (plus VAT). Enterprise and Private Cloud options are available by separate agreement.

Yes. Every seat currently includes €10 in AI credits per billing month (the price list governs). It is used before purchased credits; unused monthly allowance expires at month end and doesn’t carry over.

Prepaid usage units for AI consumption. Purchasing additional credits carries a 7.5% service fee, shown at checkout and non-refundable.

Not while you have a paid plan. Only if the account stays 36 months without a plan and without usage do remaining credits expire — with reminders roughly 60 and 14 days beforehand; any login resets the clock.

Yes. On termination, we refund purchased remaining credits automatically and regardless of amount to your last-used payment method — no request needed. Only if that fails technically (e.g., expired card) do we ask for a bank-transfer request. Free credits and the service fee are non-refundable.

Any time in text form to legal@zentre.ai (additionally — where offered — via the admin console); we confirm receipt. Monthly plans end at the end of the current billing month, annual plans at the end of the term. If you terminate because of a change we initiated (e.g., a price increase), prepaid fees are refunded pro rata.

Yes, for Enterprise customers by agreement; the prepaid expiry and refund rules then don’t apply.

3. Your data, in plain language

With Scaleway (France), Paris region; processing takes place within the EU/EEA. AI requests run through the EU gateway Cortecs exclusively to endpoints in the EU/EEA — in both model classes (“EU-hosted” and “EU-sovereign”).
By us: never. With the integrated model providers, training and fine-tuning on your content are contractually excluded — contractual integration including zero data retention is a precondition for an endpoint being activated at all.
Your AI content and usage data stay in the EU/EEA; both model classes require endpoints in the EU/EEA. Only ancillary services may involve third countries: e-mail delivery (Infomaniak, Switzerland — adequacy decision) and payment processing (Stripe, USA possible — EU-US DPF/Standard Contractual Clauses). Remote support is performed by our own, bound employees; there is no third-country access to Professional Secrecy Mode content.

Chat histories are retained for 90 days and then deleted automatically. Your knowledge base remains until you delete it or the contract ends.

Export first, deletion second: you can export your data; deletion starts no earlier than 31 days after the contract ends and never before a requested export has been provided.

Yes. Export of your knowledge base (original formats), chats (JSON, within retention), and configuration — on request to legal@zentre.ai, usually within 14 days and free of charge. Details: “Switching & Portability” page in the Trust Center.

Yes, per Art. 28 GDPR, with annexes on processing, technical and organizational measures (TOM), subprocessors, and retention. It is incorporated with the main agreement.

For platform data you are the controller and we are the processor: we support you with export and deletion functions; requests that reach us directly are forwarded to you.

4. Security

TLS 1.2 or higher in transit, AES-256 (at minimum) at rest.

We operate an ISMS aligned to ISO/IEC 27001; certification is underway.

Through logical tenant separation with technical access controls. In Professional Secrecy Mode you receive a dedicated, isolated tenant with its own knowledge base.

Yes — single sign-on (SSO) including multi-factor authentication.

We notify affected customers without undue delay, at the latest within 48 hours of becoming aware of a personal-data breach — with the information you need for your Art. 33 GDPR assessment.

Hosting: Scaleway (EU). CDN: Bunny (EU delivery). AI gateway: Cortecs (EU) — model providers are engaged through it. Web search: Linkup (EU). Plus e-mail (Infomaniak), billing (Stripe), consent (Usercentrics), analytics (Plausible), bot protection for signup and login (Friendly Captcha, EU). The current list is in the Trust Center.

At least 14 days in advance by e-mail to your admin address; the Trust Center list is updated in addition. Endpoints of new model operators in Professional Secrecy Mode are likewise announced at least 14 days before activation; a 30-day period applies to the replacement of already-approved models there. In case of a justified objection, a special termination right per the DPA applies.

5. Contracts & legal

The Terms (B2B) and the DPA; for professional secrecy holders additionally the § 203 confidentiality agreement, and for Enterprise, DORA/NIS2 addenda where applicable.

The version shown at contract conclusion (identified by its date). You receive all accepted documents as PDFs with your order confirmation; previous versions remain available in the Trust Center.

German. English versions are for information only.

Electronically. The DPA applies with the main agreement; the § 203 agreement and the DORA/NIS2 addenda are each concluded separately.

German law; place of jurisdiction is Mainz.

6. Regulated sectors & EU AI Act

Yes — via the Professional Secrecy Mode (in Germany also “§ 203 mode”) together with a separate confidentiality agreement. It covers the confidentiality duties in Germany (Sec. 203 Criminal Code), Austria (e.g., Sec. 9 RAO, Sec. 80 WTBG 2017), and Switzerland (Art. 321 Criminal Code). A standard DPA alone is not sufficient for professional secrets.

A configuration for protected secrets: dedicated, isolated tenant with its own knowledge base, EU hosting, exclusively the EU endpoints authorised for the model class you selected (Annex 2 of the confidentiality agreement), zero data retention, and web search disabled.
In Professional Secrecy Mode you select one of two model classes. Both cumulatively require: inference exclusively via EU endpoints, zero data retention, no training on your content, and a confidentiality undertaking from every link in the chain. In the “EU-sovereign” class, in addition, no link in the processing chain — hosting, content delivery, AI gateway and the operators of the endpoints — is subject to the law of a third country. The “EU-hosted” class additionally includes frontier models via EU endpoints; there, the operator or its ultimate parent may additionally be subject to third-country law. You make the selection per tenant; it is logged and can be changed at any time.

Yes — for Germany (Sec. 203 Criminal Code), Austria (professional and criminal law, e.g., Sec. 9 RAO, Sec. 80 WTBG 2017, Sec. 122 Austrian Criminal Code), and Switzerland (Art. 321 Criminal Code), with the statutory instruction and a flow-down of duties to subprocessors. Profession-specific variants (BRAO, StBerG, WPO, BNotO, PAO, healthcare, WTBG-AT) and the statutory annex are in the Trust Center; the EU gateway Cortecs has confirmed a corresponding additional undertaking.

In Professional Secrecy Mode you select one of two model classes. In the “EU-sovereign” class, no link in the processing chain — hosting, content delivery, AI gateway and the operators of the endpoints — is subject to the law of a third country; the risk of foreign disclosure obligations (e.g., the US CLOUD Act) is thereby structurally minimised. The “EU-hosted” class additionally offers frontier models via EU endpoints; a residual risk remains with the operator of the endpoint. In both classes: processing exclusively in the EU, zero data retention, no training on your content, every link in the chain bound to confidentiality. Via the allowlist you control which models your team uses.

Yes. For financial entities (DORA) and NIS2 entities, addenda are available under our Enterprise agreements — they don’t apply automatically but are concluded separately. Talk to us: legal@zentre.ai.

Less than many fear: for normal office use of AI (texts, research, internal documents), you face hardly any duties of your own. EZTO handles the technical transparency duties as provider; what remains with you as deployer are a few clearly defined points — see the following questions.
The transparency duties under Art. 50 EU AI Act (Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744) apply from 2 August 2026. EZTO as provider fulfils the duty under Art. 50(1): it is clearly recognizable in the application that you are interacting with an AI system. The machine-readable marking of AI-generated content under Art. 50(2) applies to the service only from 2 December 2026 pursuant to Art. 111(4) AI Act — it is nevertheless already active today; EZTO implements it voluntarily ahead of the statutory start date. As deployer, you remain responsible for your duties under Art. 50(4) — e.g., disclosing AI-generated text you publish to inform the public.
Internally: no. If you publish AI-generated text to inform the public on matters of public interest, you must disclose the AI generation — unless the content has undergone human editorial review and responsibility. The machine-readable marking in generated files is already provided by Zentre today (legally required for the service only from 2 December 2026, Art. 111(4) AI Act).
Yes — the AI-literacy duty (Art. 4 EU AI Act) has applied to deployers since February 2025: employees using AI should sufficiently understand how it works, its opportunities and limits. A short, documented briefing is usually enough for office use cases.

No. Zentre is a platform for accessing general-purpose AI models and is not intended for high-risk uses within the meaning of Annex III of the AI Act — such as candidate selection, creditworthiness assessment, or the administration of justice. Such uses are excluded under our Terms. Anyone who uses or modifies the Service such that it constitutes a high-risk AI system thereby becomes the provider of that system within the meaning of the AI Act and assumes the corresponding obligations (Art. 25 AI Act). Your own work as a lawyer, tax advisor, or doctor is not affected: research, drafts, and submissions for your clients — including in connection with court or arbitration proceedings — remain permitted. What is excluded is use by or on behalf of a judicial authority or dispute resolution body in the adjudication itself.

Yes — the transparency features (AI notice, marking in generated files and exports) work identically in all modes and do not process any content outside your tenant to do so.