This week, the UK government moved from AI defence strategy to operational delivery with the launch of Taskforce RAID, while the EU tabled landmark legislation to lock US firms out of defence-grade AI procurement. On the battlefield, Eurosatory 2026 in Paris showcased a new generation of AI-autonomous ground systems, signalling that the shift from pilot to deployed capability is accelerating across NATO allies.
Top story: The UK Prime Minister launched Taskforce RAID at London Tech Week, a new military unit with powers to bypass standard procurement rules and fast-track AI into frontline armed forces.
UK Launches Taskforce RAID to Fast-Track Military AI Frontline
GOV.UK · Strategy
Prime Minister Keir Starmer announced the Rapid AI Delivery Taskforce (RAID) during London Tech Week, a new unit tasked with getting AI-enabled tools into the hands of British Armed Forces at speed. The taskforce reports directly to the Chief of Defence Staff and has been granted exemptions from standard financial and procedural controls to move at the pace of technology rather than conventional defence acquisition. Its initial priorities include AI-powered intelligence fusion, faster operational decision-making, and integrating autonomous systems to reduce risk to human life — though industry insiders warn that underlying MoD procurement culture will ultimately determine whether RAID delivers or stalls.
EU Moves to Bar US Firms From Defence AI Procurement
Grosswald · Regulation
The European Commission presented its Cloud and AI Development Act on 3 June as part of a Technological Sovereignty Package, establishing a four-tier sovereignty framework for public-sector AI and cloud procurement. The top tier — covering defence — would effectively require EU-made hardware and software, locking out non-European vendors such as Palantir and major US hyperscalers. The move follows news that Germany’s Bundeswehr had already decided against using Palantir for its military cloud project, and signals a structural shift in how Europe intends to control the AI supply chain underpinning its armed forces.
Eurosatory 2026: AI-Autonomous Ground Combat Systems Go on Show
UK Defence Journal · Generative AI
Eurosatory 2026 in Paris this week became a showcase for battlefield AI, with Leonardo’s IDV subsidiary unveiling the CL2X — a 16-tonne hybrid uncrewed light tank designed to act as an AI-guided robotic wingman alongside crewed main battle tanks. Separately, Ondas launched its ‘Autonomy at First Contact’ architecture, integrating air defence, ground robotics, loitering munitions, and AI command software into a single autonomous system of systems. The concentration of autonomous combat platforms at one event reflects how rapidly AI-driven autonomy is shifting from research concept to procurement-ready hardware across European defence markets.
Britain Awards Direct Contract for AI Intelligence Fusion Platform
UK Defence Journal · Tools
The UK Ministry of Defence has moved to directly award a contract worth just under £227,000 to London-based Defence Holdings PLC for ‘Project STRONG’ — a prototype AI platform designed to fuse intelligence data into a single analytical environment, generate courses of action, and support rapid, human-controlled deployment of cyber and information effects. The contract bypasses competitive tendering under Schedule 5 of the Procurement Act 2023, using provisions for prototype development. While modest in value, it represents a concrete example of the UK government beginning to operationalise its AI-in-defence ambitions rather than merely strategise about them.
Nation-State Actors Deploy LLM-Powered Adaptive Malware in Active Operations
Cyber Defense Magazine · Risk
New reporting confirms that nation-state and government-backed threat actors have begun deploying malware that uses large language models during execution, enabling attack code to generate scripts, modify itself to evade detection, and create malicious functions on demand. This marks a significant escalation beyond earlier proof-of-concept research, as AI-enabled malware can now alter attack behaviour mid-execution without requiring a human operator. For defence and security practitioners, this development compresses response windows dramatically — CrowdStrike data shows AI-enabled attacks now achieve a median network breakout time of just 29 minutes.