AI for European Organizations With Sensitive Data

For organizations and teams that work with highly sensitive data, such as trade secrets or personal data. These are usually NIS-2-regulated organizations in critical sectors, for example energy, IT and telecommunications, finance and insurance, healthcare, and government and public administration – with or without formal critical infrastructure status. It also includes teams that handle sensitive data every day, such as finance, HR or leadership teams.

Within the organization, AI Workspace meets the needs of four roles:

• AI and digital transformation leaders: roll out AI across the organization without risks blocking the rollout.

• Security, compliance and data protection: keep control over data flows and prove compliance at any time.

• IT and platform teams: get a scalable infrastructure instead of scattered API keys.

• Finance and procurement: get transparent costs and predictable budgets instead of many individual subscriptions.

Employees work in the chat, and development teams integrate AI directly into their own applications.

If you use an LLM directly – that is, a provider's AI language model via its web interface or API – you manage access yourself. Each person or department has its own accounts. This quickly becomes confusing and is hard to control.

AI Workspace brings all AI requests together in one place. The same rules apply to every use, regardless of provider or model: unified login, central cost control, protection of sensitive content and complete logging. This prevents uncontrolled AI use ("shadow IT") and keeps all usage visible.

No. You freely choose which model handles each request. AI Workspace automatically identifies the right provider from the model and forwards the request – no complex switching required. You can also set central rules to automatically redirect certain requests to a different model or a backup provider.

You specify the model you want. AI Workspace automatically sends the request to the right provider. Admins can also set fixed rules, for example to distribute requests evenly or to use a specific model by default.

If a provider goes down or is overloaded, AI Workspace automatically forwards the request to a predefined backup provider. Operations continue. If needed, AI Workspace distributes requests across several providers. A provider with a persistent outage is temporarily removed from routing until it is reliably available again.

Exception: If the error is clearly caused by the request itself, for example invalid input, AI Workspace does not switch to a backup provider.

Yes. In addition to cloud providers, you can integrate models that you host yourself, for example on your own servers. This works via Ollama, for instance, or with other models that support the widely used OpenAI standard.

Yes. BYOK stands for "Bring Your Own Key". You store the access keys of the AI providers you already have contracts with in AI Workspace. AI Workspace uses these keys to communicate with the providers on your behalf.

Your benefit: Existing contracts, terms and billing stay the same. You keep full control over your keys and can replace or revoke them at any time.

AI Workspace offers two levels of protection for sensitive data.

For highly sensitive data and trade secrets, you use open-weight models hosted on Myra-owned certified EU infrastructure. The data never leaves Myra's infrastructure, and third parties have no access.

If you want to work with frontier models, PII masking protects personal and other sensitive data. Before a request reaches the provider, automatic masking detects details such as names, addresses or credit card numbers and makes them unrecognizable. If a request violates your policies, it is blocked. To do this, every request goes through a fast basic check and a more in-depth, language-based check, both on Myra's infrastructure. Each user decides which information receives special protection.

Admins can mandate which models are used for the organization, individual teams or users. The colour coding of the chat interface shows users at all times which mode they are working in.

No. Myra Security does not use your data to train AI models – our terms of use make this binding. Your data also stays protected with the models of external providers in AI Workspace, in different ways depending on the model. As a general rule:

Open-weight models from Myra are hosted on Myra-owned EU infrastructure. Your data stays with Myra and does not reach third parties.

Models from external providers: AI Workspace accesses them via the providers' business API. For this access, most providers exclude in their terms the use of your data to train their models. Where a provider could in principle use data for training, AI Workspace prevents data collection with technical measures, as far as the provider's interface allows.

Bring Your Own Key: You use your own contract with a provider. The terms of that contract then apply. Please check there whether training on your data is excluded.

Yes. AI Workspace detects typical manipulation attempts and blocks them automatically. These include:

• Jailbreaks: attempts to make the AI bypass its security rules.

• Prompt injection: hidden instructions intended to secretly change the AI's behaviour.

Yes. Your access keys are stored in a secure key vault and encrypted with AES-256, a recognized encryption standard. This keeps them protected while stored.

AI Workspace is hosted on Myra-owned certified EU infrastructure. Myra holds several independent credentials:

• ISO 27001 based on BSI IT-Grundschutz

• BSI C5 Type 2 for cloud security

• PCI DSS Level 1 for the secure handling of payment card data

• IDW PS 951 Type 2

• KRITIS qualification under the BSI Act

• Trusted Cloud

• ISO 9001

These credentials confirm information security from infrastructure to data processing. All services are operated from German data centers in compliance with the GDPR.

Myra is a German company with 100% European ownership and operates exclusively under European law. When you use open-weight models on Myra's infrastructure, there is no access under the US CLOUD Act. Myra is trusted by DKB, 1&1 Versatel, Aleph Alpha, Munich Security Conference, EDEKA and many more.

Yes. You can allow access to AI Workspace only from approved IP addresses, for example only from your company network. This prevents unauthorized access from outside. IPv4 addresses are currently supported; IPv6 is currently not supported.

Yes. You decide how long conversations (saved chat histories) are kept and when they are deleted. You can export individual conversations at any time to back them up or reuse them.

Yes. You can add files directly to the chat via drag-and-drop and have the AI process them. AI Workspace supports images, PDF documents, plain text files, Word documents, and spreadsheets, among other formats. A progress bar keeps you informed of the processing status.

Yes. Describe what you want to see, and AI Workspace creates a matching image – right in the same interface, without an additional program. This is useful for first drafts, illustrations or quickly visualizing an idea.

Yes. You can share conversations with colleagues so they can continue working in the existing chat – as a shared conversation or as an exported file.

Yes. Using MCP connectors, you can integrate additional tools and data sources so that the AI can, for example, access other systems. You can set up these connections individually or for the entire team; AI Workspace securely manages the associated access credentials as well.

Yes. If you send exactly the same request again, AI Workspace returns the stored response from the cache without querying the AI provider again. This saves costs and time.

• The cache only applies to identical requests. If any detail changes, the request counts as new.

• It is activated per gateway and strictly separated by area and gateway. A stored response never reaches another area.

• The cache does not apply to responses output live (streaming).

Yes. AI Workspace uses the same interface as OpenAI. Existing applications therefore usually work without changes. Which provider responds is determined automatically by the selected model.

Yes. You can also connect tools like Claude Code. A documented guide is available.

Yes. With MCP connectors, you connect additional tools and data sources so the AI can access other systems. You set up these connections for yourself or for your team. AI Workspace manages the credentials securely.

With personal access keys (tokens). Each person manages their own keys and sets usage and cost limits per key.

Yes. You can create separate areas, for example for departments, teams or clients. Each area has its own budgets, rules, access and analytics. The groups use the same platform without affecting each other or seeing each other's data. This makes AI Workspace multi-tenant capable.

The price for AI Workspace is based on three factors: the number of seats, the selected platform and the model setup.

The number of seats can be set flexibly, from a single workplace to a company-wide license.

For the platform, you choose between two options depending on your data sovereignty requirements. The EU-located infrastructure offers hosting in Europe. The EU-governed infrastructure is the Myra Cloud: hosted on Myra-owned certified EU infrastructure, with data residency in the EU and locally operated LLMs.

For the model setup, three options are available. You can use more than 20 AI providers as well as open-weight models hosted on Myra-owned EU infrastructure. In this case, Myra bills the token costs directly, and the amount depends on the models your teams use. With "Bring your own key", you connect your own provider's key. Token usage is then billed through your existing contract, and no token costs are charged via Myra. A mixed setup combines both options and is priced individually.

Yes. You set cost limits on three levels: per access key, per area and for the entire AI Workspace. Limits apply per day, per month or as a permanent overall limit. Quotas reset automatically once the defined period ends. When a limit is reached, AI Workspace rejects further requests with a clear notice. This prevents unexpected costs.

Yes. You define how many requests are allowed in a given period – in total and per access key. The limit applies before a request is sent to the AI provider.

A dashboard shows usage and costs, broken down by area, provider, model and person. A live view shows current requests and detailed reports.

To calculate costs correctly, you enter the prices of the models. For models without a stored price, AI Workspace does not show any costs.

Yes. AI Workspace logs every request in full: which provider and model were used, which area the request belongs to, its size, the costs incurred and whether protection filters were triggered. This makes every use traceable, for example for internal audits and compliance.

Yes. AI Workspace automatically forwards logs to common SIEM systems, including Splunk, Elasticsearch, OpenSearch, Vector and Syslog. This way, events flow directly into your existing security monitoring.

Yes. In addition to the live view, there is a status overview of availability and metrics that you can integrate into common monitoring tools. This helps you detect issues early.

AI adoption platforms provide a chat and project interface that helps organizations roll out AI widely. Functionally, they are similar to AI Workspace. The difference lies in data security and sovereignty: Where is the platform operated? Who is involved in processing? Which independent credentials are in place?

AI Workspace is hosted on Myra-owned certified EU infrastructure. Myra is a German company with 100% European ownership. The infrastructure is certified to ISO 27001 based on BSI IT-Grundschutz and holds BSI C5 Type 2, among others. On top of that, Myra has provided security expertise for critical infrastructure since 2012.

AI gateways control the traffic between applications and AI providers. They provide routing, cost control and protection rules and are aimed mainly at development teams. What they lack is an interface where business teams actually work with AI. This has to be built, operated and connected separately.

AI Workspace natively combines both: the governance of an AI gateway and an interface for all employees. The interface is kept simple and modelled on familiar tools, so even users with little AI experience find their way immediately. One provider, one contract, one operating model – instead of an integration project. And everything is hosted on Myra-owned certified EU infrastructure.