News Briefing 14 July 2025

Editorial: 

Rapid advances in artificial intelligence are rewriting the calculus of security and defence, and this briefing offers a stark reminder that governance has failed to keep pace. From Ukraine’s battlefields, the Henry Jackson School warns that quad-copter swarms costing ‘one-thousand dollars’ can now spot, pursue and destroy multi-million-dollar assets with ‘minimal human input.’ Russia’s MS001 drone, described by its designers as a ‘digital predator,’ shows how commercially available chips let machines see, decide and strike without external commands. In parallel, industry partnerships—such as RTX’s decision to fuse Shield AI’s Hivemind into operational weapons—normalise ‘human-machine teaming’ in which pilots set objectives and code executes lethal tasks.

These developments magnify long-standing legal and ethical gaps. A United Nations Regional Information Center for Western Europe piece cautions that autonomous weapons already ‘operate with limited or no human intervention,’ while international law still ‘grapples to keep pace.’ European defence dossiers underline the strategic asymmetry that cheap autonomous munitions create when they outmatch legacy armour; experts fear that, in the absence of enforceable rules, algorithmic brinkmanship could trigger unintended escalation. Scholars further highlight emergent vulnerabilities: poisoned training data that silently re-targets humans, and soaring energy demands that may dictate who can field ‘agentic’ forces.

The window to act is narrowing. Unless states translate recognition of risk into codified restraint, tomorrow’s conflicts may be prosecuted at machine speed, beyond human judgment and outside existing accountability regimes.

Autonomy in weapons systems, and AI in security and defence

• International: Cheap Drones, Expensive Lessons: Ethics, Innovation, and Regulation of Autonomous Weapon Systems: This brief by the Henry Jackson School of International Studies warns that ‘swarming technologies, which use large numbers of low-cost autonomous systems,’ erode traditional cost-exchange advantages and make expensive platforms ‘vulnerable to $1 000 quad-copters.’ It argues the Ukraine war shows how AI-enabled loitering munitions can ‘select, pursue and strike targets with minimal human input,’ amplifying the risk of escalation and civilian harm. The author calls for a legally binding treaty that preserves ‘meaningful human control’ and urges governments to invest in counter-drone defences rather than ever-higher autonomy.
• International: Committing to the responsible use of artificial intelligence in the international sphere: The UN Regional Information Center for Western Europe article states that ‘AI is already reshaping the landscape of modern conflict.’ Autonomous Weapons Systems ‘increasingly operate with limited or no human intervention,’ raising accountability and escalation concerns. It laments an emerging ‘regulatory void’ and insists on ‘a proactive, inclusive and rules-based approach’ that embeds human rights and ‘meaningful human control’ at the centre of any global compact.
• International:How International Law Is Adapting to Emerging and Disruptive Technologies: ‘Traditional frameworks of international law grapple to keep pace amidst an era of rapid technological advancement,’ the piece notes, listing AI and autonomous weapons as foremost challenges. It stresses that the ‘accountability of an AI system for making a lethal decision on a battlefield is still a debate,’ and warns that reactive law-making ‘could prove perilous.’ The author urges forward-looking treaties that clarify liability, tighten export controls and guarantee human oversight before machines may apply lethal force. 
• International: Security and Defence: The dossier introduces a series on Europe’s re-armament and cautions: ‘Drones are now an integral part of the modern battlefield and fully autonomous weapon systems are on the way, although regulations are still lacking.’ It highlights the asymmetry created when cheap AI-guided munitions outmatch legacy armour, underscoring civil-society demands for stronger export rules and a new international instrument to reinstate human decision-making in targeting.
• International: RTX, Shield AI Partner to Add Autonomy, AI to Defense Systems: The Defense Post reports that RTX will embed Shield AI’s Hivemind autonomy stack into an ‘operational weapon,’ the first system powered by Networked Collaborative Autonomy. Executives say the deal enables ‘human-machine teaming, where humans set objectives and the AI pilot executes them autonomously.’ While marketed as efficiency, the project exemplifies rapid AI-targeting capability and heightens concerns that lethal decisions could migrate from pilots to code, reinforcing calls for treaty clauses that require a human to remain ‘in, on or above the loop.’
• Russia: Russia field-testing new AI drone powered by Nvidia’s Jetson Orin  supercomputer: Major-General Klochkov reports the MS001 ‘is able to see, analyze, decide, and strike without external commands… It is a digital predator.’ Using a Jetson Orin module, the drone can reprioritise targets mid-flight and swarm with peers, defeating GPS jamming. The episode shows both the global diffusion of high-end AI chips and the urgent ‘risks associated with autonomous weapon systems’ that can pursue humans without supervision—exactly the scenario campaigners cite when demanding a binding ban
• Qatar/ International: Killer algorithms: When code becomes a tool of extermination: ‘Automated systems now make life-and-death decisions without direct human oversight,’ this piece warns, calling this shift ‘algorithmic deterrence.’ It asks, ‘What happens if rival algorithms clash on the battlefield? Could a war erupt due to a computational error?’ and stresses the absence of any treaty compelling transparency or human control. The author argues that unchecked AI targeting humans introduces an era where ‘the democratisation of digital killing’ places civilians and non-state actors inside a lethal autonomous loop.
• France: A game changer for France’s military: How AI is transforming the defence industry: The report from France24 notes that France plans to invest €2 billion in defence AI by 2030. French forces are already trialling predictive maintenance, autonomous drone swarms and decision-support software. While framed as strategic modernisation, the investment underscores broader European support for autonomy in weapons systems and intensifies pressure for international rules ensuring robust ‘human-in-the-loop’ safeguards before lethal deployment. 
• International/ US: Data Poisoning as a Covert Weapon: Securing U.S. Military Superiority in AI-Driven Warfare: This piece highlights that though ‘rapid integration of artificial intelligence… has revolutionized modern warfare,’ it also opens a new vulnerability: corrupted training data. The author contends covert U.S. data-poisoning campaigns could ‘achieve a decisive asymmetric advantage,’ while acknowledging a reciprocal threat. By showing how AI can be silently redirected to mis-target humans, the piece illustrates systemic risks and buttresses treaty advocates who want verification regimes and auditability to keep autonomous systems under meaningful human control. 
• India: The new battlefield: AI-based warfare in the ‘agentic’ age, multi-domain ops and energy as a big constraining factor: Indian Express reports that even before China’s DeepSeek model, its PLA was ‘integrating generative AI with military drones to automatically target opponents’ radars,’ the article reports. Experts warn India that ‘modern armed forces must analyse large volumes of data from even more domains… to decode enemy movements,’ yet soaring energy demands for AI could become a strategic bottleneck. The piece frames autonomy as inevitable but insists on clear doctrines that retain human decision-makers, echoing coalition messaging on responsible control.
• US: Defense Department budget request goes hard on AI, autonomy: Defense One reports that ‘the new Defense Department budget request seeks big investments in artificial intelligence and autonomy,’ including $789 million for collaborative combat aircraft and new drone-swarm lines. Whether the Pentagon can buy fast enough to match AI’s pace is questioned, but the scale underscores a shift toward systems where ‘wingman’ drones may select and engage targets. 
• US: $10M Deal Accelerates AI Deployment Across Pentagon Command Centers: The Pentagon has ‘awarded Ask Sage $10 million to expand the use of large language models across all Combatant Commands,’ aiming to streamline planning, logistics and even weapons development. Embedding generative-AI advisers deep in command-and-control chains may speed decisions but also raises fresh reliability and accountability questions—particularly if AI outputs influence lethal orders. The story highlights a fast-closing window to secure governance mechanisms that keep humans firmly in charge. 

AI, algorithms and autonomy in the wider world

• China/ International: Experts Demand Human Control Over AI in Critical Decisions: At Beijing’s 13th World Peace Forum, security specialists warned that ‘life-and-death decisions must remain exclusively under human authority, never delegated to autonomous systems.’ Speakers from the ICRC and EU diplomats stressed that no computer will ever bear responsibility for civilian harm and urged a treaty to guarantee ‘meaningful human control’ over all AI weapons. The panel highlighted legal ambiguity when AI-driven drones mis-identify targets, calling such black-box decisions a direct threat to international humanitarian law.
• International: Letter: Asimov’s assertions on AI were never more relevant: This Financial Times letter invokes Isaac Asimov’s first law—robots must not harm humans—to argue that present AI development risks overturning that principle. Pointing to rapid moves toward autonomy in weapons, the author insists that international norms must embed human decision-makers ‘at the heart of the AI story’ and draws a parallel with nuclear-era arms control: delay in regulation today could prove catastrophic tomorrow.
• International: The Ethical Dilemmas of Emerging Technology: A McLean Forrester Perspective: McLean Forrester warns that ‘unchecked AI development poses significant risks’ ranging from bias and deep-fakes to ‘autonomous weapons [that] could destabilise global security.’ The essay calls for mandatory transparency, third-party audits and ethical guidelines, stressing that without robust oversight AI may ‘exacerbate inequality and erode trust.’ It ties the debate to human-control language, arguing that only accountable governance can keep innovation from sliding into harmful autonomy.
• International: OpenAI’s security crackdown signals new AI battle: In this piece by TechCrunch, OpenAI has imposed ‘information tenting’ rules that limit staff access to sensitive models and now isolates core algorithms on offline machines protected by biometrics. The company accelerated the clamp-down after alleging that a rival Chinese lab copied its systems via ‘distillation’ techniques. The article frames the move as an escalating contest for strategic AI capabilities—a reminder that powerful models, if diverted to military use, could enable targeting of humans without consent, underscoring calls for strict external controls. 
• International: Warfare rewired: drones and civilian risk: Human Rights Watch’s Belkis Wille describes a ‘paradigm shift’ in conflict: cheap commercial drones now let non-state actors ‘target civilians in a highly precise manner.’ She warns that AI-enabled autonomy could soon create ‘killer robots… with no human in the loop,’ capable of striking a child as easily as a tank. The ICRC notes that existing law still applies but enforcement is weak; experts therefore call for stronger export controls and a new treaty before drones evolve into fully autonomous weapon systems.
• International: Hired by an Algorithm: The Unchecked Power of AI in Recruitment: This ORF commentary says AI tools are reshaping hiring, yet their ‘unregulated use raises urgent concerns about bias, accountability, and data privacy.’ Algorithms that screen résumés or analyse video interviews can entrench discrimination and make opaque decisions that ‘no human recruiter can easily audit.’ Although outside the battlefield, the piece echoes Stop Killer Robots’ messaging: when algorithms determine people’s futures—employment or survival—society must insist on transparency and human oversight.

Facial recognition, biometric identification, surveillance

• Kazakhstan: Kazakhstan bans face coverings that impede facial recognition in public: Biometric Update reports that President Kassym-Jomart Tokayev has signed an amendment to Kazakhstan’s ‘Law on the Prevention of Offences’ prohibiting masks or garments that ‘interfere with facial recognition’ in public spaces. Officials argue the measure will ‘increase the effectiveness of crime prevention’ and bolster counter-extremism operations, while permitting medical, occupational or weather-related coverings. Critics warn the law joins a regional trend toward restricting religious dress and expanding state surveillance. The move signals Kazakhstan’s determination to optimize nationwide facial-recognition deployments despite civil-liberty concerns.
• UK: Corsight’s crisis: why British police forces must rethink their Israeli facial recognition partners: Action on Armed Violence’s investigation urges UK police forces to reconsider contracts with Israel-based Corsight AI after ‘mounting investor concerns over alleged human-rights violations linked to its technology in Gaza.’ Author Iain Overton notes that Corsight markets near-real-time face-matching for law-enforcement, yet critics say its algorithms have been field-tested on occupied populations, raising ethical and legal red flags. The article contends that partnering with such vendors risks legitimising ‘opaque surveillance systems whose provenance is mired in conflict,’ and calls for rigorous human-rights due-diligence before further UK procurement. 
• US: Facial recognition systems based on older methods show high levels of bias: Study: Reporting on a U.S. government-backed study, Biometric Update states that legacy open-source detectors such as OpenCV’s Haar Cascades miss faces with darker skin tones 24.3 percent of the time, versus 0.28 percent for the lightest tones. Researchers attribute the disparity to outdated training data and urge removal of ‘poorly performing algorithms’ from public-facing deployments. Lead author John Howard cautions that continued reliance on these models ‘shows almost 10× levels of skin-tone bias,’ underscoring the need for updated benchmarks and mandatory bias audits before commercial or governmental use.

Reports:

• Report: Second-order impacts of civil artificial intelligence regulation on defense: Why the national security community must engage: The Atlantic Council report ‘Second-Order Impacts of Civil Artificial Intelligence Regulation on Defense’ warns that civilian AI rules can inadvertently constrain or shape military autonomy. It argues that dual-use AI blurs existing carve-outs and may limit the ‘tools available to the defense and national-security community,’ expose commanders to novel liability, and raise the cost and risk of fielding systems that can autonomously target people. The author urges defense actors to engage early in drafting ‘legally binding’ standards, align with civilian technical norms, and insist on persistent human control and assurance mechanisms to mitigate escalation and compliance risks linked to autonomous capabilities.