New initiative to strengthen Swedish cybersecurity
As part of the efforts to bolster Sweden’s operational readiness against cyber threats, AI Sweden is announcing the next phase of its national initiatives for AI security. In collaboration with the Army Staff, the National Cyber Security Centre (NCSC-SE), and the Näringsvärnet, which includes several of AI Sweden’s partner organizations, the initiative will develop the capability to rapidly identify and manage threats against operations involving critical infrastructure or highly sensitive systems. Simultaneously, the project will test Sweden’s mobilization capability in practice.
"Sweden must immediately prepare for the launch of highly advanced, automated cyberattack tools," says Robert Bridges, Mathematician & Innovation Leader at AI Sweden.
The purpose of this new initiative is twofold: To build the capacity to quickly identify, test, prioritize, and strengthen the systems and operations that are most vulnerable or most critical to society, and to test Sweden’s mobilization capability in practice to understand how well we can assemble the right actors, create a shared direction, share responsibility, and maintain momentum when new, potentially grave threats emerge.
The fact that we can gather leading experts from AI Sweden and our partners and collaborate with authorities like the NCSC and the Army Staff so quickly demonstrates the strength of the Swedish ecosystem. When we join forces across the public sector, the business community, and expert environments, we can both build actual capability and show that Sweden can mobilize rapidly when new threats arise. We can be certain that the development of similar AI tools is happening in many places around the world right now
Mats Nordlund
Director AI Labs at AI Sweden.
Sweden must quickly prepare
On April 7, 2026, Anthropic announced Claude Mythos Preview, an AI model capable of independently identifying and exploiting software vulnerabilities. Within a few weeks, the model discovered thousands of critical vulnerabilities, including in several of the world’s most hardened operating systems; some of these flaws had remained unknown for over two decades.
The development of advanced AI models, with Mythos being one of the latest, means that new weaknesses can be identified faster and that cyberattacks can increasingly be automated, accelerated, and executed with higher precision than before. This shifts the landscape for both defense and readiness. Particularly vulnerable are operations where the consequences of intrusion, manipulation, or service disruptions are severe, such as public administration, defense, healthcare, critical infrastructure, and other sensitive environments.
Sweden must quickly prepare for the imminent launches of highly advanced, automated cyberattack agents. Mythos Preview, from Anthropic, has demonstrated significant offensive cyber capabilities. Among other things, it can independently exploit vulnerabilities it has found in all major operating systems, and it has also shown a capacity for autonomous subversive activity.
Robert Bridges
Mathematician & Innovation Leader at AI Sweden
Briefing on the Implications of Mythos Preview and Project Glasswing for Sweden
In collaboration with the Army Staff, the NCSC, and the National Business Defense, AI Sweden invited decision-makers to a briefing on Friday, April 24. The briefing, which was held digitally and was open to everyone, brought together experts and representatives from the Swedish business community and several of the involved government agencies. Watch the recording:
Related content
AI and Cyber Resilience
Understanding AI security: Machine learning system vulnerabilities
Secure AI
Follow us
AI Sweden Newsletter
Sign up for the latest in Swedish AI
AI Sweden is formally hosted by Lindholmen Science Park AB.