The more capable artificial intelligence becomes, the more important one question becomes: how do we make sure it does what it is supposed to do — and nothing it shouldn't?
For Michael Nilsson, an engineer educated at Chalmers University of Technology in Sweden, this challenge had a familiar foundation. Over many years, Michael had developed an engineering methodology for managing complexity, maintaining control over execution and turning intentions into verifiable outcomes.
When AI began moving towards greater autonomy, Michael and entrepreneur Peter Swartling recognized an opportunity to apply these principles to an entirely new challenge. They believed that making AI more intelligent was only part of the equation. Making that intelligence trustworthy, controllable and useful in the real world was equally important.
Around two years ago, they began developing what would become HexaBox. Not another AI model. A foundation for putting AI to work responsibly.
Different backgrounds. A shared ambition.
Peter brought decades of entrepreneurial experience, beginning with a technology venture and his first exit in the 1990s. His subsequent career in the music industry and as a business leader had given him a particular appreciation for bringing together talent, ideas and the right people to make things happen.
As the underlying methodology and intellectual property took shape, Jimmy Lindberg joined the team. A former professional musician who moved into digital technology in the early 1990s, Jimmy had spent much of his career leading businesses, developing digital solutions and managing complex technology projects. His focus was on translating the emerging architecture into something tangible: a working system capable of delivering the value the inventions promised.
Together, the three perspectives — engineering, entrepreneurship and practical technology development — shaped HexaBox into what it is today.
From principles to a working platform.
The work evolved into an integrated architecture built around three connected capabilities. The Agent Factory creates and configures specialized AI agents. The Control Plane governs how those agents operate, with the declared mandate, approval requirements and governance requirements built into the execution framework and every decision recorded. And the Driver Seat, with its visual Wind Shield, gives people an intuitive way to direct, observe and understand what their AI agents are doing.
The underlying approach is simple: control should not be something added to AI afterwards. It should be part of how the system works from the beginning.
This philosophy has also guided the company's growing intellectual property portfolio — thirty invention families, patent-pending, priority date November 2025 — including complementary inventions addressing practical challenges encountered along the way.
Putting it to the test.
Building the technology was one milestone. Demonstrating that it could work in practice was another.
Together with 7Security, HexaBox developed Eihwaz, an AI-driven cyber-security fleet built on the platform. Using the Agent Factory, the team created and customized a fleet of specialized agents, integrated them into the application and tested the complete solution in a real-world proof of concept. It has run in production through a managed security provider since.
The result was a functioning product and an important validation of the approach. It also provided something equally valuable: practical experience that continues to inform the development of the platform.
The journey ahead.
In April 2026, HexaBox AB was formally established as a privately held company in Stockholm, with an expanding team and a long-term strategy centered on making advanced AI capabilities accessible through partners and specialized applications. Behind the founders sit a chairman, Stefan Thorberg, and a deputy board member, Johan Staël von Holstein, who have built and sold companies before.
Our ambition extends beyond any single industry. As AI becomes increasingly autonomous, its potential will reach into almost every part of society. In highly regulated environments, the need for security, accountability and traceability is already obvious. Elsewhere, the consequences of uncontrolled actions may be less apparent, but no less significant.
We believe organizations should be able to benefit from increasingly capable AI without losing the ability to understand, govern and take responsibility for its actions. There is still much to develop, learn and prove. But the direction is clear.
We are building the control that makes greater autonomy possible.
And we believe that is an essential part of making AI genuinely useful — for businesses, people and society.
HexaBox AB · Stockholm · Org. nr 559587-2739 · contact@hexabox.one · Trust & privacy