This week’s defence AI landscape is defined by a tension between expanding capability and contested governance: the EU AI Act entered full enforcement, the Responsible AI Defense Act cleared a key Senate committee, and researchers confirmed that autonomous AI attack tools have crossed from theory into demonstrated reality. Meanwhile, Europe is accelerating physical AI integration — from next-gen AI tanks showcased at Eurosatory to the UK government launching a national cyber shield programme — signalling that both offensive and defensive AI are rapidly moving out of the lab and into operational deployment.
Top story: Researchers from Toronto, Cambridge and the Vector Institute demonstrated a self-replicating AI worm that autonomously exploits networks using free open-weight models — confirming that autonomous cyberoffence is no longer theoretical.
Self-Replicating AI Worm Proves Autonomous Cyberoffence Is Now Real
Help Net Security · Risk
Researchers at the University of Toronto, Vector Institute, and University of Cambridge built and tested a proof-of-concept AI-driven worm that does not rely on a fixed exploit list — instead it analyses each target, reasons about how to attack it, and creates a bespoke strategy on the fly using a small, free LLM running on machines it has already compromised. In tests it exploited an average of 23 hosts per run across a mixed Linux, Windows and IoT network with no attacker involvement after initial deployment. The researchers concluded that ‘autonomous cyberoffence has crossed from theoretical risk to demonstrated capability’ — a direct challenge to defence and critical national infrastructure operators whose patch cycles cannot match machine-speed adversaries.
https://www.helpnetsecurity.com/2026/06/03/autonomous-ai-worm-prototype/
Eurosatory 2026: AI Moves Inside the Tank as Europe Rearms Armour
Army Recognition · Strategy
At Eurosatory 2026, every major European armoured vehicle manufacturer showcased AI as a core system rather than an add-on, with platforms featuring AI-assisted decision-making, advanced sensor fusion, and native integration with unmanned systems to transform tanks into command-and-combat nodes within a broader battlefield network. European defence firm KNDS unveiled the Capint, an advanced armoured combat system built around AI capabilities and next-generation battlefield technologies, reflecting a post-Ukraine shift in which AI-enabled targeting, drone coordination and real-time data sharing are now considered as critical as firepower. The convergence of surging European defence budgets and battlefield lessons from Ukraine is accelerating procurement timelines, with the Franco-German Main Ground Combat System and Italy’s I-MBT representing the most ambitious long-term AI armour programmes on the continent.
UK Launches National Cyber Shield With £90M SME Investment Call
Computer Weekly · Strategy
UK Security Minister Dan Jarvis has called on leading AI companies to work directly with the government to build a ‘national cyber shield’ — defensive AI capable of identifying and repairing software vulnerabilities at machine speed — after NCSC-managed security incidents more than doubled in 2025, driven by hostile state AI-enabled attacks. The government simultaneously committed £90 million to secure small and medium-sized businesses, which it identified as the most exposed segment of the national supply chain, and wrote to over 180 leading UK businesses urging sign-up to a new Cyber Resilience Pledge. The initiative signals a structural shift in UK defence policy: rather than purely reactive cyber response, the government is explicitly building autonomous AI-powered defence infrastructure and treating frontier AI firms as sovereign security partners.