News briefing 28 July 2025

Editorial: 

The AI ‘race’ has moved from research labs to doctrine. The United States’ new AI Action Plan proclaims ‘a race to achieve global dominance in artificial intelligence,’ arguing that ‘whoever has the largest AI ecosystem will set global AI standards.’ It tasks the Pentagon to refine Responsible AI frameworks, create an ‘AI & Autonomous Systems Virtual Proving Ground,’ and seed war colleges as AI hubs—while conceding generative AI’s ‘lack of predictability.’ Capability is outpacing governance, and speed is being treated as a virtue.

Operational use is already here. At Nellis AFB, software ‘reduced cognitive load’ and issued ‘real‑time recommendations for sensor‑to‑shooter decision chains,’ with airmen nominally in charge. Germany’s push—’money is no longer an excuse’—accelerates bio‑robotics and micro‑drones touted as ‘as revolutionary as the first tanks.’ Israel frames AI as a strategic edge across surveillance, targeting, and mission execution. Diffusion is rapid: Turkey’s Roketsan fields a weaponized quadruped blending ISR and strike; China’s live‑fire show unveiled MALE UAVs, loitering munitions, and a vehicle‑mounted laser for counter‑UAS operations.

The accountability scaffolding is thin. Reporting from recent conflicts indicates AI tools ‘make very fast recommendations’ reviewed by humans for ‘about 20 seconds,’ scaling strikes from ’50 to 100 targets a day.’ This risks turning human judgment into ‘rubber‑stamping.’ Analysts warn of governance lag, ‘black box’ decisions, and automation bias. Legal scholars highlight the drift from human‑in‑the‑loop to on‑ and off‑the‑loop control, complicating proportionality and diffusing responsibility among programmers, manufacturers, and commanders. The ICRC’s standard—systems that are ‘predictable, reliable, understandable and explainable’—is far from guaranteed under current practices.

This is a global stability problem. Voices from the Global South caution that governance gaps make regions ‘unintended testing grounds,’ while proposing a convention to ban fully autonomous weapons. Meanwhile, great‑power rhetoric around ecosystems and dominance compresses crisis decision time and fuels action–reaction cycles.

Autonomy in weapons systems, and AI in security and defence

US: How the Pentagon plays into Trump’s sprawling artificial intelligence ‘Action Plan’: In this piece by Breaking Defense, White House’s America’s AI Action Plan declares ‘The United States is in a race to achieve global dominance in artificial intelligence,’ and ‘Whoever has the largest AI ecosystem will set global AI standards.’ The plan lists ‘about 19 ‘recommended policy actions’’ involving DoD, spanning ‘AI Interpretability, Control, & Robustness,’ ‘AI Hackathons,’ and standards for ‘high-security AI centers.’ It asks DoD to ‘continue to refine DOD’s Responsible AI and Generative AI Frameworks,’ proposes an ‘AI & Autonomous Systems Virtual Proving Ground,’ and makes war colleges ‘hubs of AI research,’ noting generative AI’s ‘lack of predictability.’

US: Air Force experiments with using AI to seek combat targets: Air Force Times describes a Nellis AFB trial in which software ‘reduce[d] cognitive load’ and ‘offered real‑time recommendations for sensor‑to‑shooter decision chains,’ while airmen stayed responsible for choices. Leaders said they are ‘not just testing software’ but ‘how teams make decisions with AI.’ The report underscores USAF’s aim for ‘resilient, data‑driven’ kill webs and human‑on‑the‑loop control as targeting grows more complex. 

International: Spy cockroaches and AI robots: Germany plots the future of warfare: Reuters profiles Germany’s defense push into AI‑enabled ground robots and micro‑drones. A government commissioner says, ‘Money is no longer an excuse,’ signalling funding resolve. One firm claims its bio‑robots are ‘equipped with thermal cameras, fiber‑optic sensors and night vision,’ and a defense official calls their impact ‘as revolutionary as the first tanks.’ The piece highlights dual‑use tech, swarm tactics and European rearmament dynamics alongside ethical and legal debates.

International: Project Ploughshares researcher calls for ‘meaningful human control’ over AI-enabled warfare: Branka Marijan writes for Anglican Journal and argues AI systems in targeting ‘make very fast recommendations to humans who spend about 20 seconds reviewing,’ and in Gaza the shift meant ’50 to 100 targets a day.’ She warns humans often become ‘just rubber‑stamping,’ so policy must ensure ‘meaningful human control’ yielding accountability. She notes some states prefer ‘appropriate levels of human involvement,’ a framing that could permit ‘minimal’ involvement. Ultimately, ‘the ultimate decision to engage… must remain entirely under human control,’ to avoid scaled errors and civilian harm.

International: Military AI and the void of accountability: The brief by Digital Watch Observatory warns ‘military AI is advancing faster than global rules can keep up,’ creating ‘urgent questions about accountability’ and ‘black box’ decisions. It cites examples from ‘Project Maven,’ ‘AI‑powered targeting in Gaza,’ and ‘semi‑autonomous drones’ in Ukraine to argue risk of ‘over‑reliance’ and ‘automation bias.’ The piece stresses that ‘meaningful human control… lacks clarity,’ leaving responsibility ‘murky’—whether it should rest with ‘the operator, the developer, or the machine itself.’ With a US‑China AI race, governance ‘lags far behind,’ raising escalation concerns. 

International: Analysing AI Warfare Through Realist Critique of International Law: The analysis contends AI‑enabled drones can ‘identify and engage targets without real‑time human control,’ making proportionality and accountability ‘complicated.’ It distinguishes ‘humans in‑the‑loop,’ ‘on‑the‑loop,’ and ‘off‑the‑loop,’ warning the latter is of ‘great concern.’ Current law ‘assume[s] human decision‑makers,’ while AI may ‘shield’ accountability among ‘manufacturer’ or ‘programmer.’ The ICRC’s view that weapons ‘must be ‘predictable, reliable, understandable and explainable’’ underpins calls for more human control. Absent rules, autonomy aligns with ‘power maximisation,’ reinforcing realist dynamics and arms‑race incentives.

International: The Global South’s AI moment: This piece in Pearls and Irritations argues the West’s ‘focus on achieving military superiority’ sidelines governance, while the UN has warned that missing ‘binding legal frameworks for lethal autonomous weapons systems (LAWS)’ risks ‘irreversible consequences.’ The Global South has become an ‘unintended testing ground for emerging military technologies,’ yet can lead by proposing ‘a draft convention banning fully autonomous weapons.’ If it ‘acts now,’ AI could ‘bridge divides instead of deepening them,’ shifting emphasis toward civilian benefits and multipolar regulation rather than militarized deployment.

International: Israel’s strategic edge in the age of AI & autonomous warfare: The article in Jerusalem post asserts that ‘Israel’s pioneering integration of artificial intelligence has rewritten the rules of modern warfare,’ noting the IDF ‘recognized how it could redefine military operations and smart combat years ago.’ Framed as a strategic edge, the coverage highlights rapid adoption of AI‑enabled capabilities in operational decision‑making and autonomy. While largely affirmative in tone, it underscores the scope and tempo AI affords to surveillance, targeting, and mission execution across domains. 

International: Advanced weapon systems make debut in China: Reporting from a Baotou live‑fire show, the piece by Asia News Network says Norinco ‘displayed several new or upgraded types of military drones and counter‑UAV weapons.’ The BZK‑005E MALE UAV is ‘able to operate in different weather conditions,’ carrying ‘370 kilograms of mission payloads,’ and undertaking ‘surveillance and reconnaissance.’ Norinco also showed the FL‑30 loitering munition for ‘beyond‑visual‑range… precision strike missions,’ and an OW5 vehicle‑mounted laser said to ‘spot and track targets rapidly and accurately,’ reflecting growing counter‑UAV and swarm capabilities.

International: Roketsan Blends Surveillance and Strike Capabilities in One ‘Robotic Dog’: The Defense Post report Roketsan’s quadruped KOZ is described as an unmanned platform that ‘can carry four guided munitions into high‑risk areas.’ It ‘operates autonomously or via remote control,’ with ‘a two‑hour mission endurance,’ and a design to remain stable across rough terrain. The concept merges ISR and strike roles in a compact, expendable ground system—illustrating the diffusion of autonomy from air to ground and the convergence of sensing and effect delivery in small, semi‑autonomous platforms.

AI, algorithms and autonomy in the wider world

International: Can AI really code? Study maps the roadblocks to autonomous software engineering: MIT News summarizes a CSAIL agenda: despite progress, ‘there’s also a long way to go’ to autonomous software engineering. The authors argue headline metrics miss real‑world complexity: ‘code completion is the easy part; the hard part is everything else,’ from large‑scale refactors to testing and maintenance. They call for richer datasets, better benchmarks, and tools that expose model uncertainty to keep humans in control. 

International: Experts react: What Trump’s new AI Action Plan means for tech, energy, the economy, and more: Summarizing the White House pitch as ‘an industrial revolution, an information revolution, and a renaissance—all at once,’ contributors for this piece in Atlantic Council argue the plan emphasizes ‘full‑stack AI export packages,’ yet is ‘all sticks, few carrots’ on partnerships. Analysts caution the US ‘can’t lead the way on AI through dominance alone,’ stressing trust and co‑designed norms. Others question the ‘AI race’ framing and warn implementation details remain thin amid cuts to tech diplomacy and standards‑engagement capacity.

Facial recognition, biometric identification, surveillance

US: Unlawful use of facial recognition by police erodes public trust in biometrics: This piece by biometric update reports on an NYPD case, the article says police ‘outsourced a Clearview search through a fire marshal’ despite policy limits—identifying a protester from ‘high school formal’ and ‘graduation’ photos. A judge wrote the emails ‘contravene official NYPD policy,’ and dismissed the case. The piece argues that when authorities ‘break the rules… a pillar of public trust… gains yet another crack,’ noting transparency orders and broader legal headwinds around Clearview’s law‑enforcement use in the US.

China: China Uncertainty Leads to New Christian Crackdown, ‘Face Recognition Cameras’ in Churches: In this piece by CBN, interviewees describe heightened surveillance: ‘All four corners of the church, including the pulpit, have to install face recognition cameras to monitor all those who are attending.’ The report alleges arrests over ‘tithes and offerings,’ and claims party‑aligned controls extend to worship practices. While framed from a Christian‑advocacy perspective, the central assertion remains the expansion of biometric monitoring in religious venues amid a broader crackdown.

International: Legal Brief: Jailbreak Reignites Facial Recognition Discussion: This piece by Security Info Watch reports that following an Orleans Parish jailbreak, police tapped Project N.O.L.A., ‘a robust network of approximately 5,000 private surveillance cameras’ of which ‘approximately 200 have advanced facial recognition capabilities.’ AI ‘alerts’ were relayed to police, ‘leading to the apprehension of one’ fugitive. The column flags tensions with a city ordinance and the core challenge of ‘balancing effective public safety… with protecting citizen privacy,’ including whether private networks acting with police become state agents under law.

International: Commentary: AI must assist police officers, not replace them: This commentary in The Post and Courier argues that AI should enhance policing while safeguarding civil liberties. It warns of ‘AI surveillance networks with the power to scan, identify and flag individuals in public and private spaces,’ noting New Orleans police ‘secretly tapped into a privately run network of 200 cameras equipped with facial recognition software.’ Risks include an ‘algorithm-driven dragnet with an imperfect accuracy record,’ which ‘should not be used as the basis for an arrest.’ Any real‑time system ‘should require at least reasonable suspicion, if not a warrant,’ and ‘during protests… it shouldn’t be used at all.’ Constructively, AI can assist with ‘the review of police body camera footage,’ helping analyze vast video archives.

Reports

US: White House — America’s AI Action Plan: The plan argues the U.S. must ‘build and maintain vast AI infrastructure,’ pursue a ‘worker‑first AI agenda,’ and for defense ‘establish an AI & Autonomous Systems Virtual Proving Ground.’ It seeks lighter regulation to ‘remove red tape,’ expand NIST‑led standards, and create ‘high‑security AI data centers’ for sensitive workloads—positioning AI as a strategic industrial and national‑security priority.

US: Acquiring Generative Artificial Intelligence to Improve U.S. Department of Defense Influence Activities: RAND finds ‘Generative AI presents opportunities for scaling and automation’ in analysis, planning and assessment for influence missions, but urges rapid, coherent acquisition: ‘DoD needs to rapidly acquire and employ’ capabilities and move beyond ‘ad hoc, bottom‑up efforts.’ Recommendations include defining requirements, leveraging enterprise infrastructure (e.g., CDAO), and establishing guardrails for AI‑generated content.