Security
Security & Data Sovereignty Architecture
Sovereign by design. Engineered for the modern enterprise.
At Attanda, security and data sovereignty are not post deployment add ons. They are the foundational layer of the managed agentic infrastructure we build and operate for you. Systems are deployed under clear architectural separation, deterministic governance and European data protection law, and this page states what that means in practice rather than in principle.
Regional data sovereignty (Frankfurt)
Application runtime, model inference, the database and every generated report run in Google Cloud's Frankfurt region, europe-west3. Reports are written to private Cloud Storage in the same region.
Elara, the assistant, is a Google Customer Engagement Suite agent deployed in the EU region.
Processing is carried out under the EU GDPR (DSGVO). The assistant's interface is loaded from Google's global CDN and the site uses analytics provided through our network provider; neither carries your Blueprint answers or your Readiness result. The Privacy Policy sets out exactly what each one collects.
Deterministic governance, not a black box
Readiness scoring is deterministic and rule based. The same answers always produce the same score. The language model writes the supporting narrative; it does not decide the result.
Every assessment is written to an immutable report snapshot before it is shown to you, so the document you receive is the document that was generated, and a later change to your Blueprint cannot silently rewrite it.
Where a deployment acts on one of your systems, the approval step is defined with you during the Blueprint rather than assumed. Human review points are a design decision we make together, not a default we apply on your behalf.
No model training on your data, and project isolation
Your data is not used to train models. Attanda does not train, fine tune or improve public or third party foundational models on customer data, and the Vertex AI models used do not train on it either.
What reaches the model is sanitised first. The evaluation payload carries your architecture choices and nothing else: no name, no email address, no contact record. Your contact details are stored separately and are never passed to the evaluator.
Multi tenant deployments use strict logical separation. Dedicated isolation for regulated accounts is scoped per engagement rather than offered as a standing configuration.
Encryption and access control
In transit: all external and service to service communication is encrypted with TLS.
At rest: everything stored in Google Cloud, including the database, report documents and logs, is encrypted by the platform with AES-256.
Report access is token based and the token is never stored. Only a hash of it is kept, so a copy of the database cannot reconstruct a working report link, and an expired or altered link returns nothing rather than revealing that the report exists.
Enterprise Data Processing Agreements (DPA)
We execute formal, legally binding Data Processing Agreements compliant with Article 28 of the EU GDPR with all corporate partners. Custom enterprise security annexes and vendor risk assessments are available on request during the Strategic Audit phase.
Security contact and vulnerability inquiries
If your IT security committee or CISO requires technical documentation or an architecture review, contact our security team directly at [email protected].
Attanda e.U., Wienerbergstraße 9, 1100 Vienna, Austria.
This page describes the architecture as it is built and operated today. It is not a certification, an audit report or a legal guarantee, and the exact safeguards for any deployment depend on the systems it connects to. Those are agreed per project and recorded in the DPA.

