Security is part of the architecture — how Responza protects your data
When a platform becomes central to your organisation's knowledge, data security is no longer a detail. It becomes part of the foundation.
This is especially true when the platform also uses AI to find, explain, summarise and formulate answers based on the organisation's knowledge.
If employees, managers, IT organisations and information security officers are to trust a knowledge platform, they must also be able to trust how data is processed, protected and used.
Data security therefore cannot be something added on afterwards. It has to be built in from the start.
It is in Responza.
Responza is built on principles such as security by design, privacy by design, least privilege and defense in depth. In practice, this means we work systematically to limit which data we process, who can access it, and how data is protected throughout the entire chain.
Not only in the platform itself. But also in the AI layer, the development process, operations and the way access is managed.
Data security starts with the architecture
When talking about data security, it is easy to focus on individual elements. A firewall. An encryption standard. A supplier agreement. A certification. An AI model.
But real data security is not about a single layer. It is about the architecture as a whole.
At Responza, we work on the principle that security must be built into the platform from the start. This means security is not just something checked at the end, but something that shapes design, access control, data processing, development and operations.
| Principle | What does it mean in practice? |
|---|---|
| Security by design | Security is built into the architecture from the start |
| Privacy by design | Data protection is a fundamental part of the solution |
| Least privilege | Users and systems only get the access they need |
| Defense in depth | Data is protected through multiple layers, not just one security measure |
| Data minimisation | The platform does not process more information than necessary |
Our security work is based on recognised frameworks such as OWASP and CIS, among others. Our information security management is also designed with ISO/IEC 27001 as a reference.
This does not mean security is reduced to a document or a checklist. It means we work with security as an ongoing discipline.
We do not process more personal data than necessary
A central principle in Responza is data minimisation. This means the platform should, as a rule, only process the information necessary to deliver the function. The standard platform features typically need a limited set of user information. For example:
- name
- username
- role
- access rights
- usage and statistics data
This information is used, among other things, for access control, user administration, statistics and platform operations.
The point is simple: Responza must not collect more than is needed.
Data is stored and deleted according to established principles, and access to information is managed based on the user's role and work needs.
The same principle applies in development and testing. Production data and development data are kept separate. As a rule, production data must not be used in development and test environments without controlled anonymisation or masking and relevant approval.
This reduces the risk of sensitive or business-critical information being unnecessarily exposed.
The access the user has follows them all the way — also into AI
An important difference between Responza's AI solutions and many standalone AI tools is that Responza does not establish a parallel AI universe with its own copy of the organisation's rights and data.
The AI features build on top of Responza's existing access control.
This means the access a user has in Responza also applies when the user uses AI features.
If a user does not have access to a particular content area in Responza, the AI layer must not be able to use that content as the basis for an answer either.
This is crucial. Because AI must not become a shortcut around the organisation's governance.
| Situation | What should happen? |
|---|---|
| The user has access to the content | AI can use the content as a relevant basis |
| The user does not have access to the content | AI must not use the content in the answer |
| Content is restricted to specific roles | The restriction must be respected in the AI layer |
| AI is used by different departments | The answer basis must follow the user's access |
| The organisation uses SSO and roles | Access control must align with the existing structure |
Access can be managed via, for example, SAML/SSO, roles and — for AI features — dedicated Entra ID roles.
This means AI becomes yet another way to use the knowledge the user is already authorised to see. Not a new channel where access rules disappear.
Customer knowledge is not sent uncontrolled to an AI model
When organisations consider AI, a very natural question often arises: What actually happens to our knowledge when AI needs to use it?
The short answer is that Responza's AI solutions are built on a controlled process.
Responza uses a RAG process, where AI does not simply answer freely from general knowledge. Before an answer is generated, Responza finds relevant parts of the approved content via Azure AI Search. The relevant context is then sent to the Answer Engine and Azure OpenAI to generate the explanation, summary or answer.
This means the AI answer is based on the knowledge the organisation has already established and governed in Responza. Not on a free chatbot dialogue without organisational context.
| Step | What happens? |
|---|---|
| 1. The user asks a question | The AI feature is activated in Responza |
| 2. Relevant knowledge is found | Azure AI Search finds relevant parts of approved content |
| 3. Context is passed on | The relevant basis is sent to the Answer Engine and Azure OpenAI |
| 4. An answer is generated | AI formulates an explanation, summary or answer |
| 5. Access and governance are respected | AI works within the user's access and the governed knowledge base |
This approach reduces the risk of AI answering based on random, uncontrolled or irrelevant knowledge.
Azure OpenAI can also be used with zero retention, depending on the specific solution and the customer's requirements. For organisations with particular data residency requirements, the solution can be configured to use models with inference in the Swedish Azure data centre, when the desired model is available for European inference.
The important principle is that the customer's knowledge is processed in a controlled way. Not as a free mass of text that is simply passed on uncritically.
Data is protected both technically and organisationally
Data security is about both technology and processes.
In Responza, data is technically protected through encryption, controlled entry points and secure handling of secrets and key material.
Data is encrypted both in transit and at rest. Communication takes place via HTTPS/TLS, while data in SQL, Cosmos DB and Blob Storage, among others, is protected by encryption at rest.
Secrets and key material are handled through dedicated secure services rather than being placed directly in the applications.
At the same time, the architecture limits the exposure of the platform itself. Traffic enters through controlled entry points such as Azure Front Door and API Management, while the underlying services do not need to be directly exposed to the internet.
| Security area | How it is handled |
|---|---|
| Data in transit | Protected via HTTPS/TLS |
| Data at rest | Encrypted in relevant data stores |
| Secrets and keys | Handled in dedicated secure services |
| Network exposure | Traffic is managed through controlled entry points |
| Underlying services | Restricted from direct internet exposure |
| Access | Managed by role, need and authorisation |
Security continues as we develop
The technical setup does not stand alone, however. Security also requires clear processes, roles and controls.
If something goes wrong, we have a process for it
A secure platform requires a secure development process.
That is why changes to Responza are handled through a controlled and traceable process in Azure DevOps — from change request and code change to review, test, build and deployment.
This means changes are not simply moved straight into production without control.
Code and third-party components are automatically checked for, among other things, known vulnerabilities and CVEs. If our pipelines find relevant security issues, the build process can be stopped, so the issue must be assessed and handled before the change can continue towards production.
Development, test and production are also technically separate environments. This is important for two reasons.
First, it reduces the risk of errors or insecure changes hitting the production environment. Second, it makes the development process more traceable. It must be possible to see what has been changed, why it was changed, who reviewed it, and how the change was tested.
No serious supplier should claim that a complex IT system can be made completely risk-free.
That is why information security is not only about prevention. It is also about detecting, responding and learning.
Responza uses audit logging, monitoring and security alerting. Security incidents are handled through a documented incident process with identification, analysis, containment, recovery and subsequent learning.
This means there is a process for what should happen if something unexpected occurs.
| Phase | Purpose |
|---|---|
| Identification | Detect and register the incident |
| Analysis | Assess scope, cause and risk |
| Containment | Limit the incident and prevent further impact |
| Recovery | Restore normal operations |
| Learning | Improve processes and controls after the incident |
Backup and Disaster Recovery are likewise part of the operating model, so data and services can be restored in the event of technical incidents.
The goal is not to promise that nothing can ever go wrong. The goal is to ensure that Responza can prevent, detect and handle incidents in a controlled way.
Data security is about the entire chain
To us, data security is not about one particular firewall, certification or AI model. It is about the entire chain.
Collecting as little as possible. Managing access consistently. Protecting data technically. Developing under controlled conditions. Monitoring operations. Having processes for incidents. And ensuring that AI does not bypass the principles the rest of the platform is built on.
That is how we believe you create a platform that employees, IT organisations and information security officers can trust.
Key takeaways
- 1Data security must be part of the architecture from the start — not something added on afterwards.
- 2AI in Responza builds on existing access control and approved knowledge, so AI does not become a shortcut around governance.
- 3Data is protected through data minimisation, encryption, controlled access paths, secure development, monitoring and incident handling.
Frequently asked questions
How does Responza protect customer data?
+
Responza works with data security as part of the platform's architecture. This includes data minimisation, access control, encryption, secure development, monitoring and incident handling. The goal is to protect data throughout the entire chain — from user access to AI features and operations.
Does Responza's AI use all data in the organisation?
+
No. Responza's AI features build on the access control already in place in the platform. If a user does not have access to a particular content area, the AI layer must not be able to use that content as the basis for an answer either.
Is customer knowledge sent uncontrolled to an AI model?
+
No. Responza's AI solutions use a controlled RAG process, where relevant approved content is first found via Azure AI Search. The relevant context is then used to generate answers, explanations or summaries. AI thus works from the organisation's governed knowledge base.
What do security by design and privacy by design mean?
+
Security by design means security is built into the solution from the start. Privacy by design means data protection and data minimisation are a fundamental part of the platform's design. Both principles are about preventing risks rather than only reacting to them afterwards.
What happens if a security incident occurs?
+
Responza has a documented process for handling security incidents. It includes identification, analysis, containment, recovery and subsequent learning. In addition, audit logging, monitoring, alerting, backup and Disaster Recovery are part of the operating model.
Carsten Steffensen
Teknologichef og partner
Carsten skriver om teknologi, produktudvikling, arkitektur og praktisk implementering af videns- og AI-løsninger. Han arbejder i krydsfeltet mellem teknisk produktudvikling, systemarkitektur og den organisatoriske virkelighed.
View author profile →Want to see how Responza works with security, access control and AI on a controlled foundation?
Book a demo and see how Responza protects your knowledge and data — also when AI becomes part of the solution.