This week’s defence AI landscape is defined by a deepening tension between speed and oversight: the US is aggressively accelerating military AI adoption while the Pentagon-Anthropic fallout reshapes the commercial AI vendor market. Meanwhile, the UK is navigating its own inflection point, with its long-delayed Defence Investment Plan expected imminently and the Chief of Defence Staff warning of the most dangerous security environment since the Cold War.
Top story: The Pentagon has formally shifted two-thirds of its AI workload away from Anthropic to OpenAI, Google, and Microsoft after a landmark dispute over autonomous weapons safeguards.
Pentagon Dumps Anthropic, Formalises Multi-Vendor AI Strategy
Crypto Briefing · Strategy
The Department of Defense has shifted at least two-thirds of its AI usage away from Anthropic following a protracted dispute over whether Claude could be deployed for fully autonomous weapons and domestic surveillance without usage restrictions. By June 2026, the DoD had formalised contracts with OpenAI, Google, and Microsoft as part of a declared multi-vendor strategy for AI across classified networks. The episode is a watershed moment for the entire defence AI industry, signalling that the Pentagon will prioritise unrestricted access over vendor ethics policies — and that commercial AI firms face a stark binary choice when contracting with the military.
Trump Orders US Military to Accelerate AI Adoption Across All Agencies
AP News / US News · Regulation
President Trump issued a sweeping national security memo directing the US military and all national security agencies to accelerate AI adoption, while directing the Pentagon to update its human-oversight policy for autonomous weapons within 90 days. The memo comes amid growing concern from military leaders and contractors about the pace of deployment outstripping governance frameworks. For defence practitioners, the directive effectively fast-tracks AI integration across the entire US national security apparatus, increasing pressure on allies — including the UK and NATO partners — to match pace.
UK Defence Investment Plan Delayed as AI Drone Spending Looms
Defence Online · Finance
The UK’s Chief of Defence Staff warned this week of the most dangerous security period since the Cold War, citing Russian airspace incursions and hybrid threats, as the government’s long-awaited Defence Investment Plan remains delayed ahead of the NATO summit in early July. The DIP is widely expected to include a significant shift toward AI, drones, and autonomous weapons, encouraging direct investment into UK firms. With the UK’s Strategic Defence Review already committing over £4 billion to autonomous systems, the DIP’s publication will be a defining moment for UK defence technology procurement.
Military AI Verification Crisis: No One Can Confirm What Models Will Do
Help Net Security · Risk
A new analysis published this week argues that the core unsolved problem in military AI is not ethics or procurement, but verification: there is currently no reliable method to confirm what a deployed AI model will actually do in a live operational context. With Anduril partnered with OpenAI, Palantir with Microsoft, and Lockheed Martin with Meta, the systems emerging from these partnerships carry a security problem that sits entirely outside existing arms control frameworks. The piece draws a direct historical analogy to Cold War arms treaty failures and argues this verification gap is the defining strategic risk of the current AI arms race.
Agentic AI Worm Compromises 73% of Simulated Networks at Infosecurity Europe
TechTimes · Risk
Researchers at the University of Toronto’s CleverHans Lab demonstrated at Infosecurity Europe 2026 in London that a free, open-weight AI worm could compromise 73.8% of a simulated enterprise network autonomously — a live proof-of-concept that dominated the three-day event. The demonstration coincided with a White House AI cybersecurity executive order and CISA advisories on active attacks against fuel infrastructure, giving the conference an unusually operational edge. A survey released alongside the event found 64% of UK cybersecurity leaders believe agentic AI will have the biggest impact on the threat landscape over the next three years, marking a clear shift in the European security community’s threat model.