News Briefing 25 August 2025

Editorial:

Across recent reporting, we see a common pattern: rapid deployment, widening theatres of use, and thin guardrails of autonomous weapon systems. China’s parade of hypersonics and autonomous platforms signals strategic coercion capability; Germany is scaling drones; the US Navy is racing to “weaponize” data; Turkey and Israel are fielding AI-enabled C2 and edge systems; Kazakhstan is institutionalising defence AI. The operational reality has arrived; the governance has not.

This diffusion brings clear risks of digital dehumanisation and error cascades if humans cannot truly understand, predict, and control system behaviour. Technical voices emphasise the need to ‘open the black box,’ while operational plans envision near-real-time software updates to battlefield algorithms. That combination—opaque models and accelerated deployment—demands verifiable auditability, reproducible testing, and command accountability before, during, and after use. ‘AI is not a turnkey solution to military dominance’; it is a complex, failure-prone stack that must remain subordinate to human judgment.

Multilaterally, optimism about AI’s economic utility has not translated into credible constraints on military use. With great-power rivalry eroding previous safeguards, the case for a legally binding instrument is stronger than ever: prohibit systems that target people and those that cannot be used with meaningful human control; regulate the rest with positive obligations for context-appropriate human judgement, rigorous test-and-evaluation, and incident reporting. As the ICRC reminds states, if you invest in defence, you must also invest in international humanitarian law (IHL).

Autonomy in weapons systems, and AI in security and defence

• China: China military parade to show off hypersonic missiles and autonomous weapons: This piece by the Financial Times reports on China’s military parade showcasing new hypersonic and uncrewed systems, signaling the PLA’s expanding capacity to target US naval forces and coerce Taiwan. Officials promised the ‘debut of hypersonic, anti-missile defence and strategic missiles to exhibit our strong strategic deterrence capabilities,’ alongside ‘new tanks, autonomous vehicles for ground, naval and air use’ and ‘directed energy weapons and electronic jamming systems.’ Rehearsals revealed missiles marked ‘YJ-15, YJ-17, YJ-19 and YJ-20,’ and a force of ‘more than 10,000 personnel, 100 aircraft and several hundred ground vehicles.’
• International: Multilateral Efforts and Military Applications of AI: This piece by Observer Research Foundation reports on the G7’s upbeat AI agenda while warning that multilateral action still trails AI militarisation. The Ottawa-led summit signaled ‘a positive shift in tone’ toward real-world applications and launched a ‘G7 GovAI Grand Challenge’, with funding for ‘AI and energy solutions’. But ‘From Gaza to the Black Sea, the use of AI on the battlefield is a lived reality’, while ‘multilateral forums have only played a limited role in substantively confronting the militarisation of AI’. The article urges endorsing efforts to ‘set clear regulations and prohibitions on autonomous weapons systems by 2026’.
• International: Without AI guardrails, sci-fi nightmares about killer robots could come true: The Globe and Mail reports on the rollback of U.S. AI safeguards, accelerating an AI arms race. Biden’s 2023 order sought ‘safe, secure and trustworthy development of AI,’ but President Trump ‘revoked the Democrat’s executive order and threw away the safeguards.’ Vice-President JD Vance has previously said that ‘the AI future is not going to be won by hand-wringing about safety.’ With the U.S. and China in ‘a new global AI arms race,’ warnings by multilateral forums and civil society that ‘we cannot delegate life-or-death decisions to machines’—and vows to ‘retain human control over autonomous weapons systems’ look precarious.
• Germany: Germany plans to acquire 8,300 military drone systems by 2029, trailing allies: This piece by The Defense Post reports on Germany’s plan to field about 8,300 unmanned systems by 2029 amid a rapidly intensifying drone arms race. The package includes around 5,700 reconnaissance drones, 560 interceptor systems, and 2,070 loitering munitions. Berlin favors flexibility over bulk buys even as the Ukraine war shows drones causing ‘roughly 70 percent of casualties.’ Industry voices stress that ‘the real innovation in drones is autonomy, but the real power comes from mass,’ while capacity builders caution, ‘We won’t use our own funds to build up capacities that will never be called upon.’
• Kazakhstan: Kazakhstan Sets Up Military AI Unit to Digitize Defense: Military AI reports on Kazakhstan establishing a defense AI unit to accelerate digitization while maintaining human-supervised decision support. The new department is described as ‘a dedicated artificial intelligence unit’ that will ‘analyze large data sets,’ ‘automate management,’ and ‘enhance decision-making’ across missions. It will also ‘train officers and specialists’ and partner with universities and research centers to build national capability. 
• International: Building Trust in Military AI Starts with Opening the Black Box: War on the Rocks reports on the need for transparency and explainability to preserve meaningful human control over military AI. The authors argue proprietary systems should be ‘open and auditable to trusted government partners’ and urge leaders to ‘resist the urge to accelerate deployment.’ They caution that ‘AI is not a turnkey solution to military dominance,’ calling instead for internal expertise, rigorous testing, and accountable procurement so commanders truly understand what systems do and when they fail.
• Türkiye: AI Integration Targets Faster, Smarter Decisions in Turkish Naval Operations: This piece by Military AI reports on AI integration in Turkey’s naval C2 to augment human operators with safeguards. HAVELSAN’s MAIN platform gives ADVENT ‘improving decision-making, situational awareness, and operational efficiency,’ runs with ‘role-based access control’ and ‘data privacy safeguards,’ and aims to ‘speed up decision-making’ while supporting collaboration between human operators and automated systems. Deployed first for maintenance assistance, the roadmap extends to real-time operational support, threat classification, and training—centering oversight over automation.
• International: Israeli Firm Brings AI Video and Battlefield Computing Solutions to London Defense Expo: Military AI reports on commercialization of battlefield AI video and edge-computing, underscoring the need for controls against digital dehumanisation. Maris-Tech will showcase ‘AI-powered video and advanced computing systems,’ including “Diamond,” offering ‘full 360-degree coverage.’ The CEO says, ‘Our solutions are built to meet the tough demands of today’s defense forces, delivering real-time intelligence directly from the field.’ Such capability highlights why verifiable human oversight and targeting constraints remain essential as sensing and fusion accelerate.
• US: US Navy Unveils Push to ‘Weaponize’ Data and AI: This piece by Military AI reports on the Navy’s plan to ‘weaponize’ data and AI—turning raw feeds into tactics and software updates fast—while confronting bottlenecks that can imperil accountability. Wagner says, ‘We need to spin this really fast,’ and admits past practice ‘treated [data] like trash.’ Priorities include end-to-end pipelines, automated classification, and ‘sandbox’ environments to cut ATO delays—steps that must embed auditable controls to prevent opaque automation from outpacing meaningful human control. Another piece in Breaking Defense reports on the US Department of the Navy’s push for rapid adaptation via ‘data and AI weaponization,’ raising governance and oversight stakes. CDAO Stuart Wagner insists, ‘We need to spin this really fast,’ describing a ‘digital OODA loop’ and lamenting that vital sensor data ‘was treated like trash.’ The plan targets automated classification and sandboxed deployment to move updates in near-real time—changes that demand robust accountability to keep humans in control of consequential decisions.
• International: ICRC president: “International humanitarian law only as strong as leaders’ will to uphold it”: ICRC reports on shoring up international humanitarian law as technology transforms warfare, emphasizing political will and accountability. President Spoljaric states IHL is ‘only as strong as leaders’ will to uphold it’ and reminds states, ‘If you invest in defence, you must budget for IHL.’ On remote and autonomous means, she cautions, ‘Distance does not absolve responsibility.’ The message aligns with ensuring human judgment and lawful use remain paramount as states modernize forces and adopt AI-enabled systems.

AI, algorithms and autonomy in the wider world

• International: We Are the AI Generation: Reflections from the United Nations (UN) AI for Good Summit: This piece by iAfrica reports on inclusive, human-centred AI leadership emerging from Africa and a call for shared responsibility in steering AI for social good. Opening the UN summit, the ITU chief declared: ‘We are the AI generation,’ a line the author frames as a “call to all of us” to guide AI “toward the common good.” The essay urges non-bloc thinking—‘Don’t look West or East — look forward’—and showcases community-rooted solutions like MamaMate and Rwanda’s Health Intelligence Centres. 
• International: The Guardian view on Britain’s AI strategy: the risk is that it is dependency dressed up in digital hype: This piece by The Guardian reports on the risk that Britain’s AI strategy amounts to ‘dependency dressed up in digital hype,’ with public data and services enriching foreign tech. The editorial warns the UK may ‘hand over its data, infrastructure and public services to US tech giants’ for notional efficiency, and cautions against becoming a site of ‘extractivism.’ It presses for digital sovereignty—public cloud, open models, and institutions that keep ‘meaningful human control’ over deployment—rather than outsourcing core capabilities.

Facial recognition, biometric identification, surveillance

• UK: Human rights regulator criticises Met’s use of facial recognition cameras: BBC News reports on the equality regulator challenging the Metropolitan Police’s live facial recognition and calling for tighter safeguards. The EHRC argues ‘the Metropolitan Police’s current policy falls short of this standard’ and will intervene in a judicial review. EHRC chief John Kirkpatrick insists: ‘There must be clear rules which guarantee that live facial recognition technology is used only where necessary, proportionate and constrained by appropriate safeguards.’ The Met responds that ‘our use of LFR is both lawful and proportionate,’ even as there is ‘no specific domestic legislation regulating police use of live facial recognition tech.’
• UK: UK govt building on live facial recognition, announces predictive policing prototype: This piece by Biometric Update reports on a government-backed ‘predictive policing project’ that will fuse ‘AI-powered interactive crime maps and facial recognition technology’ and be ‘operational across England and Wales by 2030.’ Officials tout prevention benefits and link the prototype to an ‘expanded rollout of live facial recognition vans.’ Civil-liberties concerns linger around data sharing and proportional use, raising questions about human oversight and safeguards against digital dehumanisation in public-space surveillance.
• UK: ‘Facial recognition can make mistakes, it’s not a decision-maker’: BBC News reports on the planned rollout of live facial recognition in West Yorkshire alongside assurances of safeguards and mounting rights concerns. The Home Office will deploy ’10 new LFR vans,’ and West Yorkshire Police confirmed ‘two LFR vehicles would be brought into use by the force later this year.’ Officials stress privacy controls: ‘Other faces get pixelated out automatically’ and ‘The overwhelming majority of images are deleted within 0.2 seconds.’ Critics warn the rollout is ‘without a proper legal basis’ and amounts to ‘stop and search on steroids,’ prompting calls for ‘a new legal framework.’
• US: New Orleans to vote on nation’s first real-time facial recognition policing program: This piece by Biometric Update reports on New Orleans considering the nation’s first ordinance to ‘formally authorize the use of facial recognition technology for real-time surveillance in policing.’ The system could ‘monitor public spaces live and dispatch automated alerts when an individual matches a watchlist.’ Supporters call it a ‘force-multiplier,’ while critics warn it risks a ‘citywide biometric surveillance network’ tracking residents, underscoring SKR’s cautions about rights impacts, bias, and the need for enforceable human control.