Where data is stored and where AI inference runs are two separate questions. For law firms (professional secrecy under Art. 321 StGB), wealth managers (FIDLEG, FINIG, banking secrecy under Art. 47 Banking Act / Art. 69 Financial Institutions Act), and public sector bodies (cantonal data protection law), the applicable regulatory framework determines which combination is required.

The default configuration. Documents, extracted text, and the search index are stored in an isolated account on Swiss infrastructure (Infomaniak). Only the text required to answer a specific query is sent to an AI model, and it is pseudonymized before it leaves the hosting boundary (see Data Residency in Switzerland). AI inference itself runs on infrastructure partners located in the EU/EWR, never outside Switzerland or the EU/EWR.

Data storage remains as in Section 2.1. In addition, AI inference itself — including complex, multi-step reasoning, not only short tasks — is routed exclusively to Swiss-hosted models. This tier is enabled per workspace on request to the account team; it is not a self-service toggle.

The Unplex software is installed and operated on the customer's own IT infrastructure. This is a distinct commercial and contractual model, governed by a separate On-Premise Addendum to the Unplex Terms and Conditions, rather than a configuration toggle on the standard subscription.

Under this model, the customer provides the required infrastructure (server capacity, network, operating system) and is responsible for installation, configuration, and ongoing operation. Critically, the customer connects and operates their own AI models and APIs — Unplex provides the required interfaces and documentation, but does not select, license, host, or operate the AI models in this tier. Responsibility for the quality, accuracy, and lawfulness of AI-model output, and compliance with the model provider's licence terms, rests with the customer.

Because Unplex has no access to the on-premise installation, Unplex does not process customer data under this model and is not a data processor for it; the customer is sole controller of all data in the installation. See Section 4 for support terms specific to this tier.

The self-hosted tier follows a different support model than the standard subscription, since the customer operates the infrastructure. There is no uptime/availability SLA for the on-premise installation — availability is the customer's responsibility. Unplex's support commitment covers software defects only, not infrastructure or configuration issues.

Support channels: the ticketing system at support.unplex.ai is the primary channel, available 24/7 for submission. Email is available at onprem-support@unplex.ai. Phone support (+41 44 288 37 37) is reserved for P1 incidents only, Monday–Friday 08:00–17:00 CET/CEST. Chat and on-site support are not available for this tier.

Remote access to the customer's installation is only ever provided on the customer's explicit request and approval, is logged, and is time-limited to diagnosis and resolution.

Unplex provides at least 12 months' advance notice before a software version reaches end-of-life.

Tier selection is set at the workspace level by the account team — it is not a self-service setting inside the product. To change tier, or to discuss which tier fits a specific regulatory obligation, contact the Unplex account team.

Note: this article summarizes the deployment options at a high level. For the technical detail on pseudonymization, sub-processors, and the regulatory basis for the Standard and Swiss-only tiers, see Data Residency in Switzerland and Data Security and Privacy at Unplex. For the binding legal terms of the On-Premise tier, see the On-Premise Addendum to the Unplex Terms and Conditions.