This week’s dominant theme is the acceleration of AI at the heart of national security infrastructure — from the US government deploying frontier models directly inside intelligence agencies and Air Force bases, to the UK and EU racing to build sovereign AI cyber-defence capabilities. A parallel tension is also hardening: as AI systems prove devastatingly capable in red-team drills, lawmakers on both sides of the Atlantic are pushing back with oversight legislation, while the Senate demands transparency over opaque Pentagon AI contracts.
Top story: Anthropic’s Mythos AI reportedly penetrated nearly all NSA classified systems within hours during an authorised red-team drill, triggering the first-ever US export controls applied directly to an AI software model — a precedent that is now alarming Five Eyes allies including the UK.
CISA Deploys Anthropic Mythos to Audit Government Code for Bugs
Reuters / US News · Risk
The US Cybersecurity and Infrastructure Security Agency (CISA) is using Anthropic’s restricted Mythos AI model to scan government code repositories for exploitable vulnerabilities, Reuters reported on 6 July. The deployment comes even as Anthropic navigates an ongoing standoff with the White House over military use restrictions, and follows bombshell Senate testimony that Mythos had penetrated nearly all NSA classified systems in hours during an authorised red-team drill. The episode marks the first time the US has applied export controls directly to an AI software model — a regulatory precedent that has alarmed UK and European allies who lost access with no warning.
UK Launches Agentic AI ‘Cyber Shield’ as Foreign Secretary Raises Alarm
Digit.fyi · Strategy
The UK government announced on 7 July a new initiative to integrate advanced agentic AI into its national cybersecurity framework, with GCHQ Director Anne Keast-Butler outlining a blueprint for a national cyber defence system capable of deploying frontier AI at machine speed. The announcement landed the same week that Foreign Secretary Yvette Cooper warned AI represents ‘the biggest security challenge of the decade’, calling for urgent international action to prevent AI from automating cyber-attacks and assisting in the development of biological or chemical weapons. The announcements signal a step-change in UK posture — moving from AI safety rhetoric to operational AI-enabled defence.
Pentagon Awards $80M AI Counter-Drone Contract to Protect Air Force Bases
DefenseScoop · Finance
AeroVironment has been awarded an $80.5 million task order — the first under a $500 million Army contract — to deploy its Titan-MS AI-powered multi-sensor fusion system to defend Air Force Global Strike Command bases against small drone threats. The system fuses AI and machine learning to detect, identify, track, and defeat both RF-controlled and autonomous drone threats, and is part of the Domestic Shield initiative run by Joint Interagency Task Force 401. The contract is a clear signal that AI-enabled counter-drone defence of home-soil military installations is now a top Pentagon priority, with hundreds of millions more in procurement expected to follow.
https://defensescoop.com/2026/07/06/pentagon-awards-task-order-to-av-for-titan-drone-defense/
DoD’s GenAI.mil Hits 1.7M Users, Plans Classified-Level Expansion
Nextgov/FCW · Generative AI
The Pentagon’s internal AI marketplace GenAI.mil has reached 1.7 million users and facilitated the creation of over 100,000 custom AI agents, with the Department of Defense’s chief digital officer announcing plans to push the platform to higher classification levels and add new commercial models — including ChatGPT. The ‘commercial-first’ procurement policy represents a major shift in how the US military sources AI capability, putting vendors directly alongside warfighters. Defence practitioners should note the platform is already being used to aggregate battlefield data and accelerate decision-making in ways that its own architects admit human cognition ‘is just not going to be able to keep up with’.
Senator Warren Demands Pentagon and Tech Giants Disclose Military AI Contracts
NBC News · Regulation
Senator Elizabeth Warren wrote to seven major US tech companies on 7 July demanding full disclosure of their military AI contracts with the Pentagon, citing concerns over the Defence Department’s opaque AI arrangements. Warren specifically highlighted a deal with Reflection AI, which is partially funded by a firm linked to Donald Trump Jr., raising conflict-of-interest concerns on top of existing oversight gaps. The move follows a wave of congressional scrutiny over autonomous weapons and AI procurement secrecy, and mirrors growing pressure in the UK and EU for defence AI contracts to be subject to democratic accountability.