The Noxtua Model System

Not every legal task makes the same demands. Searching across a large corpus of contracts is a different job from drafting a clause, and a filing for a public authority may come with different requirements than an internal memo. A single setup for everything works - but it means accepting the same conditions for every task, whatever you're working on.

What is the Noxtua Model System

The Noxtua Model System gives you more than one category to work with, and lets you choose between them yourself. You'll find the model picker in the chat field, and it offers different options:

Standard powered by Athena Agent - open-source/weights models, hosted exclusively on sovereign, European-controlled infrastructure.

Frontier powered by Minerva Agent - international frontier AI models (closed-source) in a secure European setup from Deutsche Telekom, under some of the highest security protections currently available.

The choice is entirely voluntary. Which categories you see depends on your market and on what your administrator has made available, and there is always a default category available to everyone - it can never be switched off.

The base models in the categories Athena Agent and Minerva Agent are combined in the Noxtua AI system with an additional set of specialist models for OCR, search, and voice which are hosted on European infrastructure.

Across every category, client data is never used to train the underlying models and is subject to zero data retention during processing, which means that the models keep nothing once a request is finished. The clients’ conversation history, which enables them to return to their work across sessions and interfaces, is stored by Noxtua in encrypted storage on European infrastructure (Deutsche Telekom and IONOS), under the Noxtua keys, and never on a model provider's systems. Clients can decide for how long this history is kept and when to delete it.

You can also learn more on our dedicated model page on our website, where the entire AI system is detailed.

Athena Agent

Noxtua is the only Legal AI that offers legal professionals in Europe a product version running exclusively on sovereign, European-controlled cloud infrastructure. Athena is that version: self-hosted open-source models on sovereign European infrastructure provided by the German cloud providers Deutsche Telekom and IONOS.

Noxtua runs local copies of the open-source/weights base model on our own infrastructure. The models are completely disconnected from the original model creator. Therefore, it is technically impossible for the original creator to access the models or any of their in- or outputs. This setup opens the possibility to use high-performant models independently from their origin (European, American, Chinese, etc.) in a sovereign setup. Athena is tailored for legal use by Noxtua.

Because Athena does not have any service providers in its processing chain that are subject to U.S. law, a disclosure request under the US CLOUD Act would have no addressee in this case. Processing occurs entirely within the European environment controlled by Noxtua. Inputs are used solely to fulfill your request, are not utilized for training the underlying models, and are subject to no data retention during processing.

This makes Athena the preferred category for the judiciary, public-sector buyers, and regulated institutions. Public authorities receive a stable version stream with a cautious update schedule and audit access.

Minerva Agent

With Minerva, Noxtua is the first Legal AI to give you access to international frontier AI models (closed-source) in a European setup with European cloud provider Deutsche Telekom under some of the highest security standards currently available. It's our answer to legal professionals who have asked for frontier model access without stepping outside that setup.

For Minerva, we work with our European partner Deutsche Telekom, which offers access to international frontier models via T Cloud Public – a specialized cloud built for high-security applications (Telekom AI Foundation Services). The closed-source model is accessed directly through T Cloud Public, over an encrypted, isolated connection into a dedicated Google processing environment in a EU data centre. Client data is processed there only for as long as the request takes, in strict isolation. Once the request is answered, nothing is retained (zero data retention). The entire processing chain is governed by a data processing agreement and complies with the professional-secrecy obligations of the jurisdictions we operate in.

With the orchestration of the security layers, this option reduces exposure to the US CLOUD Act to as far as is currently achievable when using international, closed-source frontier models. A potential US data disclosure warrant would require the disclosure of existing data a provider holds. However, the Minerva category is engineered with a zero-retention policy so that no data is held on the inference layer, while the durable store of clients’ conversation history is never on a model provider's systems but in our own Noxtua infrastructure in encrypted storage on Deutsche Telekom and IONOS, under our own keys.

Your advantages with the Noxtua Model System

The freedom to choose: You decide which category fits your requirements and your clients' - per task, and without a trade-off. All categories are fully featured in the product, so choosing one never means giving something up.

International frontier models in a European setting: Minerva gives you access to international frontier AI models with Deutsche Telekom under some of the highest security protections currently available - a first for Legal AI.

A fully sovereign option, whenever you want it: With Athena you work exclusively on sovereign, European-controlled cloud infrastructure operated by Deutsche Telekom and IONOS.

The choice sits where the responsibility sits: No one understands your requirements better than you. Administrators define by policy which categories are available in their workspace and set what everyone starts with, so the choice sits with the people who carry the compliance responsibility. A category you have selected is never overridden.

Full transparency: The origin, purpose, and operation of every category are clearly documented, and every update is published with a date.

Tested before it ships: Before a model is admitted to one of our categories, it passes pre-registered release gates - both before launch and before any modification: strict quality thresholds, legal engineering tests by C.H.BECK, and bias assessments. The models Noxtua hosts itself are fully isolated from their original providers. Updates follow a documented, rehearsed process, and a proven exit path exists for each model category.

How to use the Noxtua Model System

Step 1: Check what's available to you

Open a chat and look for the model dropdown in the chat field.

If you don't see it, the Noxtua Model System isn't available in your market – everything else in Noxtua works exactly as before.

If you do see it, the categories it lists are the ones available in your market and permitted by your administrator.

Step 2: Choose which categories your organization uses (admins only)

All available categories are switched on by default, so you only need to act if you'd rather your users didn't have all of them.

Admins can manage this under Settings → Advanced Settings. You can deselect the non-default category – for example Frontier – and your users will then work with the default category only. The default category itself cannot be deselected, so your organization can always work with Noxtua.

Step 3: Select a category

Open the dropdown in the chat field and pick the category you want to work with.

Your choice is remembered: the category you select in one chat is preselected the next time you start a new one.

Step 4: Switch whenever you need to

You're not locked in. You can switch categories at any time – including in the middle of a conversation. The new category takes over from that point on.

You can also pick up older chats with a category that wasn't available when you started them, and simply continue the conversation from there.

Step 5: Use it across Noxtua

The model picker isn't limited to the web app. You'll find it in:

• Chat – per chat, in the chat field

• Matrix – per matrix

• The Word add-in – per chat

• The Outlook add-in – per chat

Each has its own picker, so you can work with different categories in different places at the same time.