AI in security-sensitive organizations
New opportunities, on entirely different terms
AI creates new ways to analyze, compile and make information accessible at scale. For organizations with the highest security requirements, that is a clear step up in capability. At the same time, the conditions there are different from those in open environments.
AI models are not fully predictable. Results cannot be taken for granted. Information handling must be strictly controlled and auditable.
So it is not primarily about trusting the technology. It is about being able to control it.
AI for security-classified information
Built for control, not trust
When AI is used in systems that handle security-classified information, the requirements for how solutions are designed change.
Here:
- AI must be treated as a controlled component
- results must be validated before they are used
- access and information flows must be strictly limited
- behavior must be tested and verified over time
It also means AI must be usable where the data is: on your own premises, in an air-gapped environment with no internet connection, with the right protection and control.
Bring AI to the data, not the data to AI.
Craton Secure AI
Two services, one secure foundation
Craton Secure AI: Chat
Smarter casework with AI inside the protected environment.
An AI assistant along the lines of ChatGPT, Copilot or Claude, but inside your own system.
Who it’s for: Case officers, analysts, investigators and other staff who work with sensitive information every day.
What you get:
- Less time searching, more time assessing. Find, analyze and compile information from multiple documents in minutes.
- Answers you can verify. Every answer shows the source it is based on, so you know what is substantiated.
- The right information for the right person. The AI only sees what you yourself are authorized to access.
- Connect your own services. An internal API makes the AI available to other services in the system, for example SitaWare Insight.
Examples of what you can ask
“Summarize the five most recent situation reports and show where they differ.”
“Which requirements in the new regulation affect our information handling procedure?”
“Draft a report using our template, based on my notes.”
Craton Secure AI: Dev
Faster, more secure development with AI agents, without the code leaving the system.
AI agents now find and fix vulnerabilities at a pace no manual effort can match, and attackers have the same capability. Developers in closed environments have long lacked the AI tools that have become standard across the rest of the industry. Craton Secure AI: Dev gives them the same capabilities, inside the air gap and under full control.
Who it’s for: Developers, system integrators and system administrators who build, configure and maintain systems in closed environments.
What you get:
- Faster development. AI agents such as OpenCode and Claude Code help developers and administrators work faster.
- More secure code. Let the AI find vulnerabilities and suggest fixes before the code moves on.
- Your codebase stays at home. Everything happens inside the protected environment.
-
Build AI into your own services. Connect them through an internal API, inside the system.
Examples of what you can ask
> Review the module and list possible vulnerabilities with suggested fixes.
> Write unit tests for this function.
> Create a firewall rule that only allows traffic between these two services, and analyze how the change affects the rest of the system.
Frequently asked questions about Craton Secure AI
How does the approval process work?
Approval is granted for each of your systems, and we prepare the supporting documentation together with you.
Do we need Craton to use Craton Secure AI?
No. The service can be delivered separately, but it then requires a closed system with an equivalent level of protection. That is a prerequisite for us to maintain the level of security and control the solution is designed for.
Can the security function monitor how the AI is used?
Yes. The system logs who did what and which sources were used.

AI for protective security
Contact us to learn more about AI for protective security
At Basalt, we work to enable AI in environments with systems that handle security-classified information, both through our own development and in customer assignments. With Craton Secure AI, you can start now with a pre-study ahead of ordering, in which we review your needs, your environment and your security requirements.
-
Get tips and inspiration from Basalt
Subscribe to our newsletter -

Stay one step ahead.
Let Basalt be responsible for your operational protection.
Contact us and we will guide you right!
Search the page
Search for pages or news
