For our self-hosted models: Under our technical and contractual setup, the infrastructure provider has no access to client or case data, and the models run with hash-verified weights (a check confirming the model file is genuine and has not been altered), disconnected from the original creator. The models run in our environment with no outbound network path to the original creator, who therefore has no access to the models or to any inputs or outputs.

With the new NextGen, Noxtua is introducing a multi-access approach in the Minerva Agent to increase the resilience through further diversification, with models running in EU data centers. Across every category, inputs are used only to answer the request, are not used to train the underlying models, and are subject to zero data retention during processing, which means that inputs and outputs are not stored by the model provider, including for the closed-source models on all Minerva access paths. As a result, the model providers do not maintain a permanent database that could be the subject of a potential request for disclosure.

This must be distinguished from a client’s conversation history. In order for clients to be able to return to their work across sessions and interfaces, Noxtua stores their conversations in encrypted storage on our own European infrastructure (Deutsche Telekom and IONOS), and never on a model provider's systems. Clients can decide for how long this history is kept and when to delete it. Where a matter cannot accommodate even a theoretical residual risk of access by non-European authorities, Athena provides a category with no US provider in the processing chain.

The entire chain in both categories is built to satisfy the confidentiality undertakings covering lawyers’ professional secrecy.

Noxtua is certified to ISO 27001, ISO 42001, ISO 9001, ISO 27017, and ISO 27018, and holds a BSI C5 attestation, a SOC 2 report, and a TISAX label.

Last updated: 29 October 2026