This week’s defence AI landscape is defined by an accelerating push from prototype to production — with autonomous ground vehicles entering Marine Corps service and Senate lawmakers scrambling to impose governance guardrails before autonomous weapons become irreversible facts on the ground. Simultaneously, the EU’s AI Act moves into enforcement mode in August, forcing European defence and critical-infrastructure operators to grapple with binding obligations for the first time.
Top story: The US Senate Armed Services Committee voted to embed human-judgment requirements into the FY2027 defence bill, the first statutory attempt to govern autonomous weapons at scale.
Marine Corps Fields First Production Autonomous Ground Vehicles via Overland AI
GeekWire · Strategy
Seattle startup Overland AI has secured a $19.7 million Marine Corps production contract — the first ever awarded to a ground autonomy company as prime contractor — to deliver more than a dozen fully autonomous vehicles supporting the Marine Air Defense Integrated System by early 2027. Unlike remotely operated predecessors, the vehicles independently plan routes and navigate terrain after receiving a destination, removing operators from hazardous forward positions. The deal, fast-tracked through the Pentagon’s APFIT programme, marks the transition of AI-driven ground autonomy from experimentation to operational fielding and is likely to accelerate similar awards across other US military branches.
EU AI Act Enforcement Hits Defence-Adjacent Sectors From August
Bratby Law · Regulation
A detailed legal analysis confirms that from 2 August 2026 the European Commission can levy fines of up to 3% of global turnover against providers of general-purpose AI models with ‘systemic risk’ — including cybersecurity misuse risk — under the EU AI Act. The Commission’s accompanying nine-point AI-cybersecurity action plan, also published this week, layers new programmes (a European Blueprint for structured access to frontier AI, a secure ENISA testing platform for critical sectors) on top of these binding obligations. UK operators with EU market exposure face an immediate compliance decision point, particularly as the UK has no equivalent horizontal AI law and pursues a lighter-touch, regulator-led model.
AI Chat Windows Now Coordinating Jets, Drones and Missiles in Combat
The Washington Times · Generative AI
A detailed Washington Times investigation reveals that US service members are already coordinating fighter jets, rockets, and autonomous drones through single AI-powered chat interfaces, with Palantir’s Maven Smart System — now a mandatory reporting tool across multiple military branches — integrating large language models including Anthropic’s Claude for real-time intelligence analysis and battlefield scenario simulation. Defence firms including Anduril are simultaneously racing to push deployable data centres close to the front line to enable AI decision-support in denied-communication environments. The piece underscores that military AI has moved far beyond experimentation, raising urgent questions about reliability and accountability in life-or-death scenarios.
https://www.washingtontimes.com/news/2026/jul/8/combat-evolved-data-taking-warfighting-us-military/