This week’s defence AI landscape is defined by a race to embed frontier AI models into classified government infrastructure — from the Pentagon’s GenAI.mil platform to Europe’s push for sovereign open-source AI — while the EU scrambles to finalise AI Act compliance rules before an August deadline that directly affects high-risk defence and security systems. The Anthropic-DoD legal battle also reached a significant milestone, reinforcing that AI safety guardrails are now a frontline governance issue, not just a policy aspiration.
Top story: The Pentagon confirmed it is expanding its classified GenAI.mil platform with xAI’s Grok models, giving all 3 million US military and civilian personnel access to frontier AI on sensitive government networks.
Pentagon Expands GenAI.mil With xAI Grok for 3 Million Personnel
U.S. Department of War (war.gov) · Strategy
The US Department of War confirmed this week that its bespoke GenAI.mil AI platform will be expanded with xAI for Government’s suite of frontier-grade capabilities, based on the Grok family of models, targeting Impact Level 5 deployment for all military and civilian personnel. The platform is designed as a single gateway for AI tools across classification levels, with the xAI integration enabling secure handling of Controlled Unclassified Information in daily workflows. For defence practitioners, this signals a decisive shift from AI experimentation to enterprise-wide deployment — with Elon Musk’s xAI now embedded at the heart of US military operations.
Court Blocks Pentagon’s Punishment of Anthropic Over AI Guardrails
Electronic Frontier Foundation · Legal
A California federal court granted Anthropic a preliminary injunction, ruling that the US government’s designation of Anthropic as a ‘supply-chain risk to national security’ — after the firm refused to allow its Claude AI to be used for autonomous weapons and domestic mass surveillance — constituted illegal First Amendment retaliation rather than a legitimate security measure. The ruling is a landmark moment for AI governance in defence: it establishes that the government cannot punish commercial AI firms for maintaining safety guardrails. The case has become a flashpoint for the broader question of who controls the ethical limits of AI in military contexts.
EU Races to Finalise AI Act Omnibus Before August Defence Deadline
Global Policy Watch / Covington · Regulation
The EU is pressing toward formal adoption of its Digital Omnibus on AI — the first set of amendments to the EU AI Act since 2024 — with final publication in the Official Journal expected in July, ahead of the critical 2 August 2026 compliance milestone. The deal defers high-risk AI system obligations (covering law enforcement, border control, and biometric tools directly relevant to defence and security) by up to 16 months, moving the deadline to December 2027. For European defence and security organisations, this provides a narrow window of relief — but transparency obligations under Article 50 remain live from August, meaning AI-generated content in operational contexts will require disclosure from this summer.