Vaatimustenmukaisuus
Alikäsittelijät
Vaatimustenmukaisuusasiakirja
Alikäsittelijät
Tämä asiakirja on julkaistu englanniksi.
(englanninkielinen alkuperäinen)
Last reviewed: 2026-11-10 · Version: 1.10
A sub-processor is another company we engage that processes personal data on the customer's behalf. Today there are four. Three serve only our own website, its demo and its back office (the demo holds fictional data; the website's question chat is one of them; the back office at /admin holds our own sales and admin records, never a customer's workspace data); the fourth is used only when a customer enables it for its workspace. We tell customers in advance before we add or change a sub-processor of the service we deliver to them (see "Our own website" below).
| Sub-processor | Purpose | Data | Region | Safeguard |
|---|---|---|---|---|
| Vercel | Hosting this public website, its clickable demo (the demo only, not the product) and our back office at /admin (one Vercel project) | Fictional demo data and ordinary web request logs (such as IP address and browser). At /admin, what the back office handles: the sign-in of the admin accounts, the Book a call requests, contracts, install usage reports and install profiles | Washington, D.C. (United States, Vercel region iad1) for the site's server functions. The one exception is the question chat's function, which runs in Frankfurt (EU) | Per the provider's data processing terms |
| Anthropic | The AI model behind the demo's Ask bar, used when no EU model is configured for the demo | The question a visitor types about the fictional company. Please do not type personal data there | United States | Per the provider's data processing terms |
| Vercel (AI Gateway) routing to Alibaba Cloud, running an Alibaba Qwen model | The AI model behind this public website's question chat, which answers questions about the product from the website's public content only | The question a visitor types and the model's answer. Visitors are asked not to enter personal or customer data. We do not store questions or answers; we keep only a count of questions per day. Every call asks for zero data retention, and the gateway then routes it only to providers it has zero-data-retention agreements with | Any region, including outside the EU: the chat's calls are not pinned to an EU region. Every call is pinned to Alibaba Cloud only | Vercel: per its data processing terms |
| TypeSafe (Jev, hosted) | Answers typed questions (a choice, a yes/no or a score) about a record, only when the customer enables it for the workspace | The record the question is about, with the columns tagged for masking replaced by placeholders by the anonymiser before it is sent | Any region, including outside the EU, so a workspace that allows EU models only never reaches it | Data processing agreement requested |
Our own website
The three website rows above (Vercel hosting, the back office included; Anthropic for the demo's Ask bar; and Vercel's AI Gateway with Alibaba Cloud for the question chat) serve our own public website, not the service we deliver to customers (the back office at /admin is our own too). No customer's workspace data passes through them. We list them here so visitors can see who handles what they type on this website; the advance notice above applies to sub-processors of the service delivered to customers, not to these.
Why Microsoft Azure is not on this list
The product is installed in the customer's own Azure tenant, on the customer's own Microsoft agreement and bill. Microsoft is the customer's own provider under that agreement (including the Microsoft Products and Services Data Protection Addendum), not a sub-processor we engage. The same holds for AI models deployed through Azure AI Foundry in the customer's subscription.
Where an AI model processes a prompt
Where an AI model processes a prompt in the product is the customer's choice: the customer selects one of the model frameworks, and each framework states where it processes. Foundry Azure, open source (the default): outside the EU, inside Microsoft's Azure, with the default model's prompt processed in the US. Foundry Azure EU, open source: only within the EU. Frontier models in Foundry: in any Azure region, including outside the EU. Self-hosted VM in your Azure: on the customer's own machine, in its EU region. Direct access: on the model vendor's own service, outside Azure and outside the EU, billed to the customer directly.