This week’s defence AI landscape is defined by tightening governance on both sides of the Atlantic: the EU Parliament voted through the AI Omnibus that reshapes high-risk classification rules just as defence contractors scramble to comply, while NATO’s DIANA mechanism brokered its first R&D contract between an Ally and a tech company — a sign that alliance-wide AI adoption is moving from policy to procurement. Meanwhile, the Pentagon’s multi-vendor classified-network AI architecture and DIANA’s London-headquartered accelerator model are converging on the same conclusion: no single vendor should control military AI infrastructure.
Top story: The EU Parliament voted to pass the AI Omnibus on 16 June, formally delaying high-risk AI compliance deadlines to December 2027 — giving defence contractors breathing room but also crystallising a new classification regime that will govern AI in security-critical systems.
EU Parliament Passes AI Omnibus, Reshaping High-Risk Rules for Defence
Osborne Clarke · Regulation
On 16 June 2026, the European Parliament voted in favour of the AI Omnibus, with formal enactment now expected within weeks. The legislation pushes compliance deadlines for stand-alone high-risk AI systems to December 2027 and product-embedded systems to August 2028 — but the draft guidelines published alongside it adopt an expansive interpretation of what counts as ‘high-risk’, meaning more AI systems integrated into regulated defence and security products may fall under the regime than vendors previously assumed. For defence practitioners, the message is clear: the window of regulatory breathing room is real, but the classification net is wider than expected.
NATO DIANA Brokers First R&D Contract Between Ally and Tech Firm
NATO · Strategy
In April 2026, NATO’s Defence Innovation Accelerator for the North Atlantic (DIANA) — headquartered in London — activated its Rapid Adoption Service to enable the first direct R&D contract between a NATO Ally government and a technology company. DIANA focuses on deep-tech dual-use technologies including AI, autonomy, quantum and next-generation communications, running competitive challenge programmes that connect innovators with Allied military end-users. This milestone signals that NATO is moving beyond strategy documents into live procurement pipelines, compressing the timeline from innovation to battlefield deployment.
EU’s Draft High-Risk AI Guidelines Close Consultation — Defence Sector on Watch
Hunton Privacy & Cybersecurity Law Blog · Regulation
The European Commission’s draft guidelines on classifying high-risk AI systems closed their public consultation on 23 June 2026 (later extended to 23 July after stakeholder requests), with the final version expected by end of 2026. Critically for security applications, the guidelines confirm that simply including a human in the loop does not automatically declassify a high-risk AI system, and that modular or agentic AI configurations cannot be ‘carved up’ to escape classification. Defence AI developers deploying systems in areas such as biometrics, law enforcement, or critical infrastructure should treat this as a compliance trigger — not a distant concern.
Project Maven Formalised as Long-Term Pentagon Programme Under Palantir
Military.com · Finance
The US Department of Defense has formalised Palantir’s Project Maven as a permanent operational programme, signalling a shift from experimental AI tools to embedded military intelligence infrastructure. Maven uses AI to process drone, satellite and sensor imagery in near real time, feeding detections into operational planning systems — and Palantir’s data architecture connects these streams across military networks. The move reflects the Pentagon’s broader pivot toward ‘data-centric warfare’, but raises ongoing questions about the degree to which automated analysis shapes human decision-making in targeting workflows.
https://www.military.com/feature/2026/03/22/pentagon-expands-palantirs-role-ai-contract.html