This week’s defence AI news is defined by an accelerating race to embed frontier AI models into both classified military networks and national cyber defence infrastructure — with the UK, US Army, and Pentagon all making bold structural moves. Simultaneously, the EU AI Act’s full applicability deadline in August 2026 is sharpening questions about regulatory gaps covering military systems, while the GenAI.mil platform’s rapid expansion to 1.3 million users signals that AI is now a live operational tool, not a future aspiration.
Top story: The Pentagon’s GenAI.mil platform hit 1.3 million users and 57 million prompts as xAI’s Grok was formally integrated, while the US Army separately issued an RFI for agentic AI deployable on its most classified intelligence networks — JWICS and SIPRNet.
Pentagon’s GenAI.mil Hits 1.3M Users, Adds xAI Grok to Arsenal
CDO Magazine · Generative AI
The US Department of War’s GenAI.mil platform has reached 1.3 million users and processed over 57 million prompts, with xAI’s Grok models now being integrated alongside existing Google Gemini and OpenAI tools. The platform already hosts over 100,000 autonomous AI agents and has become the preferred AI system for five of six military branches. The scale and speed of this rollout — covering 3 million military and civilian personnel — marks a fundamental shift from AI experimentation to live operational dependency across the US defence enterprise.
US Army Seeks Agentic AI for Top-Secret Intelligence Networks
Intelligence Community News · Strategy
The US Army has issued a formal Request for Information for agentic AI platforms capable of deployment on its most sensitive classified networks — JWICS and SIPRNet at Impact Level 6 — with responses due July 3, 2026. The platform must enable intelligence analysts to build, deploy, and govern AI agents that can reason, retrieve information, and take actions autonomously inside accredited classified environments, including in disconnected battlefield conditions. This signals a decisive push to embed autonomous AI decision-making at the very top of the military intelligence stack, well beyond unclassified productivity tools.
https://intelligencecommunitynews.com/agentic-ai-platforms-for-army-intelligence-rfi-released/
GCHQ Blueprint Hardwires Agentic AI Into UK National Cyber Defence
Engineering and Technology Magazine · Risk
GCHQ Director Anne Keast-Butler has unveiled a blueprint for what is described as the world’s first national AI-enabled cyber defence capability, designed to use agentic AI to detect and respond to threats at machine speed across critical sectors including telecoms, airlines, and utilities. Speaking at Bletchley Park, she warned that Russia is scaling up daily hybrid activity against the UK and Europe and that cybersecurity must become ’10 times more urgent.’ The system is expected to take up to five years to fully deploy — a timeline that critics argue is dangerously slow given that the NCSC already handles approximately four nationally significant cyberattacks per week.
UK NCSC Warns AI Cyber Risk Timeline Is Now Months, Not Years
National Cyber Security Centre · Regulation
The UK’s National Cyber Security Centre has published a stark leadership briefing warning that frontier AI models are anticipated to ‘fundamentally transform both offensive and defensive cyber capabilities’ and that the critical window is ‘months, not years.’ The paper urges boards and executives to treat cyber resilience as a core business risk and to use AI deliberately to strengthen defences rather than merely improve efficiency. For defence-sector practitioners, the guidance underscores that the AI threat is no longer theoretical — adversaries are already using automated tools to find zero-day vulnerabilities faster than human teams can patch them.
https://www.ncsc.gov.uk/news/the-ai-shift-in-cyber-risk-why-leaders-must-act-now
EU AI Act’s Full Applicability Looms — But Military Systems Stay Exempt
European Commission · Regulation
The EU AI Act reaches full applicability on 2 August 2026, with transparency rules for AI-generated content also taking effect that month — yet the Act explicitly excludes technology intended exclusively for military use, leaving a significant regulatory gap as European militaries accelerate AI weapons procurement. High-risk AI rules covering biometrics, critical infrastructure, and border control will not apply until December 2027, creating a window in which dual-use defence AI sits in a legal grey zone. For European defence contractors and NATO allies, this creates both commercial opportunity and accountability risk as autonomous systems are deployed ahead of any binding governance framework.
https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai