Which subprocessors does amaise use?
amaise operates three separate environments (EU, US, CH). Services that process customer documents run in the region of the customer's environment. Supporting services (login, support, analytics, monitoring) are operated per environment or globally, as listed below; they do not receive document content.
Hosting and infrastructure:
Amazon Web Services (AWS) — Storage and compute for all customer data: documents, database, search index, backups, and e-mail delivery (Amazon SES). Processing in the environment's region: EU in Frankfurt, US in Ohio, CH in Zurich. In the EU and US, Amazon Textract additionally reads selected document pages (forms, tables, handwriting) in the same region; not used for Swiss customers. ISO 27001, SOC 1/2/3, CSA STAR, PCI DSS.
AI and text recognition (OCR):
Microsoft Azure OpenAI — LLM inference. Regional data processing in deployments restricted to the corresponding data zone (EU: EU data zone — France Central, Germany West Central, Italy North, Poland Central, Spain Central, Sweden Central, West Europe; US: US data zone — Central US, East US, East US 2, North Central US, South Central US, West US, West US 3; CH: Switzerland North exclusive). Contractually no training on customer data. Under Microsoft's terms, prompt and completion content may be retained for up to 30 days for abuse monitoring, accessible only to authorized Microsoft reviewers, and is never used to train or improve models. amaise itself logs only token counts — no prompt or completion content. ISO 27001, SOC 2.
Microsoft Azure AI Document Intelligence — OCR for Swiss customers, hosted in Switzerland (Switzerland North). Can also be enabled for EU customers (West Europe) and US customers (East US). Documents are kept only for processing and deleted afterwards (at the latest after 24 hours); no training on customer data. ISO 27001, SOC 2.
Google Document AI (Google Cloud) — OCR for EU and US customers, in the respective region: EU documents in the EU, US documents in the US. Not used for Swiss customers. Documents are stored only temporarily for processing and deleted afterwards (at the latest after two days); no training on customer data. ISO 27001, SOC 2.
Login and communication:
Auth0 (Okta) — Login, single sign-on, and multi-factor authentication. Processes user accounts (name, e-mail address, phone number for SMS codes) and login events; no document content. EU and CH: Auth0 EU region; US: Auth0 US region. SOC 2. SCCs.
Twilio (USA) — Delivery of SMS one-time codes for multi-factor authentication. Processes the user's phone number; no customer data. SCCs.
Slack (USA) — Internal team communication and operational notifications. When a user submits feedback on an AI answer, the feedback text and the user's e-mail address are forwarded to the amaise team via Slack. No document content. SCCs.
Support, analytics, and monitoring:
Intercom (USA) — In-app support chat and Help Center. Processes the signed-in user's name, e-mail address, and organization, the pages visited in the application, and the content of support conversations. Global for all environments. SCCs.
Mixpanel — Product analytics (usage behavior). Receives the signed-in user's e-mail address and name, the text entered into search and filter fields, and the questions asked of the pilots — this text can contain a person's name or e-mail address. Session replay is recorded for all customer sessions, with all text and inputs masked. No customer document content and no search results are transmitted. Data is stored in Mixpanel's EU data region for all environments. SCCs.
Sentry (USA) — Error tracking and performance monitoring, global for all environments. Error reports are configured to minimize personal data: no request bodies, no IP addresses, only the pseudonymous user ID. A sample of sessions and every session with an error is recorded as a session replay, with all text masked. No customer document content. SCCs.
Better Stack — Availability monitoring and status page. The status notice in the application is loaded from Better Stack, which receives the visitor's IP address. No customer data.
Gravatar (Automattic) (USA) — Profile pictures. The browser requests a user's picture with a one-way hash of their e-mail address. No customer data.
ipgeolocation.io — Approximate location (country, region, city) of the IP addresses shown in the audit log. Processes IP addresses only.
Optional legal research sources:
These sources are an add-on and are only used when they are activated for a workspace. They receive search queries only — no documents. Before a query is sent, the names known from the case (insured person, doctors, employer, insurer) and the case reference are removed automatically; queries that still contain a date of birth, social security number, IBAN, e-mail address, or phone number are blocked and not sent.
Perplexity (USA) — Web research. Processing in the USA. When LegalPilot is activated for a workspace, it also uses Perplexity for web research: the question entered by the user and the preceding conversation are sent as entered, without documents.
Justement (Switzerland) — Swiss case law.
Koordination Schweiz (Switzerland) — Social insurance coordination law.
entscheidsuche.ch (Switzerland) — Swiss case law.
OpenCaseLaw (Switzerland) — Swiss case law.
Fedlex (Swiss Federal Chancellery) — Federal legislation. Receives statute references only.
Zefix (Swiss commercial register) — Company lookup. Receives company names from the case (for example an employer or an insurer), because the company name is the search term.
Software development and compliance (no customer data):
GitHub — Source code and deployment pipeline.
Atlassian (Jira) — Internal task tracking.
Snyk, Semgrep, SonarQube Cloud — Analysis of source code and dependencies.
Scrut — Compliance platform for ISMS management (ISO 27001, SOC 2 lifecycle). Processes only infrastructure metadata and device compliance data; no information that could identify individual insured persons or their cases. SOC 2. SCCs.
Art. 321 Swiss Criminal Code (professional secrecy): For Swiss customers whose data is subject to professional secrecy, amaise contractually ensures that confidentiality obligations are enforced throughout the entire subprocessor chain. Core processing (documents, database) remains exclusively in Switzerland. Text recognition runs via Azure AI Document Intelligence Switzerland North. AI processing runs via Azure OpenAI Switzerland North, restricted to the Switzerland data zone; amaise logs only token counts (no prompt or completion content), and the arrangement is governed by the Microsoft Online Services DPA. Supporting services outside Switzerland do not receive document content; text that users enter themselves (search terms, pilot questions, feedback, support conversations) reaches them only as described above. The optional legal research sources and LegalPilot are only used when activated for a workspace.
Changes to the subprocessor list are communicated to customers in advance, with the right to object under GDPR Art. 28. Transfer Impact Assessments (TIA) have been conducted and documented for all US-based subprocessors.