News Briefing 30 June 2025

Editorial

Autonomy is no longer a distant military aspiration; it is landing on today’s battlefields. Ukrainian “Spiderweb” and Russian loitering drones have created a field tested precedent for low cost and mass produced automation in war. India has just field-tested an AI machine-gun that is reportedly able to identify and track targets as well as function in ‘autonomous mode’, while Saab’s Gripen E has flown dog-fight manoeuvres under a software pilot. Each such advance widens the gap between the use of lethal force and human accountability. 

That is why UN Secretary-General António Guterres urges states to ‘conclude, by 2026, a legally binding instrument’ that bans fully autonomous weapons systems and regulates all others. This deadline becomes more important as negotiations are crawling while code races ahead, and a window for preventive law is closing fast.

Yet the diplomatic clock competes with a flood of venture capital initiatives. Big Tech, fresh from cloud-computing riches, has been actively competing for their share of defense budgets: OpenAI’s $200 million Pentagon deal follows Google, Microsoft and Amazon in promising “frontier AI” for targeting, logistics and command-and-control. The Financial Times notes this turn from the sector’s 2018 anti-war stance and warns that civilian-grade software, rushed into combat, will generate errors that no one can trace—or hold accountable. 

Political theorist Dr Elke Schwarz calls this drift “the outsourcing of moral agency.” In her analysis, AI systems iterate beyond their designers, while corporate profit imperatives accelerate deployment and obscure responsibility for harm. Private platforms thus sit “between casualty figures and quarterly earnings,” distorting democratic oversight even before the application of force.

Lawmakers must therefore match technological speed with legal urgency. A robust treaty by 2026 would not halt innovation, but it would draw a bright red line: machines must not replace human decision making and accountability. Anything less invites a future where lethal power is cheap and unaccountable.

Autonomy in weapons systems, and AI in security and defence

• Taiwan/ China: Taiwan develops suicide drones akin to Ukraine’s to defend against China: Financial Times reports on ‘Overkill’, a suicide drone developed collaboratively by Taiwan’s military’s weapons research arm, and Thunder Tiger, one of the country’s leading drone makers. There is no publicly available information on autonomous features in the system, though the report mentions that the drone is powered by Auterion’s artificial intelligence strike system and camera. Auterion has also been a service provider for drones used by Ukraine’s army against Russia. 
• India: Indian Army is testing an AI machine gun that can detect, decide and destroy enemy on its own: Bharat Supply and Support, an Indian defense firm, has tested an AI powered target identification and tracking system integrated with Negev light machine guns. The system reportedly uses ‘a combination of thermal and optical sensors to detect targets in low visibility and employs ballistic compensation for factors like wind, temperature, and range.’ The report mentions the ability to control the system ‘remotely’ or in ‘autonomous mode’, the specifics of which are not elaborated further. 
• Japan: Japan Reveals New Guidelines for Incorporating AI into Defense Equipment: This piece by Asian Military Review covers Japan’s new guidelines (In Japanese) on incorporating AI into defense equipment. The guidelines has classified AI integrated defense equipment into risk categories. ‘High risk’ classified projects would have to be reviewed by a Legal and Policy Review Board to ensure compliance with IHL, and factors such as level of human involvement, human judgement and a responsible human-led command and control structure. 
• Ukraine/ Russia: Killing machines: how Russia and Ukraine’s race to perfect deadly pilotless drones could harm us all: The Guardian reports on Ukraine’s ‘Operation Spiderweb’ and similar Russian advances reveal how AI-driven drones now pick, pursue and destroy targets with no human in the loop, for as little as $10,000 – a price point that allows mass production and proliferation. Experts warn these weapons could proliferate to non-state actors, be susceptible to cyber attacks or kill unintended civilians, yet leave no one accountable. 
• International: Big Tech’s push into military AI is troubling: This report by Financial Times reports that Big Tech’s enthusiasm for military work is accelerating. OpenAI has lifted its ban on weapon-related uses and just won a $200 mn Pentagon contract to build “frontier” AI tools for war-fighting and logistics, while Google, Microsoft, Amazon, Meta and Anthropic forge partnerships with defence contractors such as Anduril. Executives sell the shift as patriotic, citing Ukraine and Israel, but critics warn code could cause lethal errors and obscure accountability for civilian deaths linked to AI-guided targeting in Gaza. 
• US: Military vets patent hallucination-resistant, explainable AI technology: Defense Scoop reports that retired U.S. veterans at startup Data² secured a patent for “hallucination-resistant,” explainable AI that bolts a knowledge-graph onto any model, tracing every answer back to original records. CEO Jon Brewton stresses that prevailing large models still hallucinate, making them “unreliable,” and in defense work “wrong answers have real consequences.” While the capability may improve trust, it highlights a bigger danger: militaries are racing to field AI whose errors could trigger deadly, potentially catastrophic miscalculations. 
• US: OpenAI wins $200m contract with US military for ‘warfighting’: The Guardian says the US Defense Department has given OpenAI a $200 million contract, its first government deal, to prototype “frontier” generative-AI for war-fighting and administrative tasks. OpenAI claims use will align with its guidelines, touting benefits for troop healthcare, cyber defences and logistics; critics see a retreat from its earlier stance against weapons work. DefenseScoop reports the agreement runs through July 2026 and that the Pentagon’s Chief Digital and AI Office expects more “frontier AI” awards soon, seeking agentic workflows for command-and-control, planning and intelligence to boost joint-force lethality and enterprise efficiency. Public transparency and congressional oversight are likely to intensify.
• International: Saab achieves AI milestone with Gripen E: Saab’s Project Beyond has flown a Gripen E fighter that hands full control to Helsing’s “Centaur” AI agent. During three test flights (May 28–June 3) the jet let Centaur autonomously run beyond-visual-range engagements, executing complex manoeuvres, fusing real-time sensor data, adapting to changing speeds, ranges and even loss of C2 links, then cueing the human pilot to fire. 

AI, algorithms and autonomy in the wider world

• International: Seeing Silicon: Can a self-driving car become a weapon?: This piece in Hindustan Times warns self-driving cars are “software on wheels”, always connected and therefore hackable. Author Shweta Taneja recalls researchers who remotely seized a Jeep in 2015 and notes today’s autonomous fleets—Waymo, Zoox, Tesla—collect GPS, biometric and payment data that could be hacked or spoofed, turning robotaxi networks into weapons. Vulnerabilities extend to third-party apps, sensor manipulation and drained batteries, fuelling an automotive-cybersecurity market forecast to jump from $3.7 bn in 2024 to $21.4 bn by 2030. Driverless convenience thus carries security trade-offs.
• International: The Need for Transparency in AI and Algorithmic Rideshare Platform Decisions: Center for Democracy and Technology (CDT) warns gig workers face severe risks when rideshare platforms let opaque AI systems autonomously decide wages, trip assignments and account status. Black-box algorithms can under-pay drivers, favour or penalise them without cause, and permanently deactivate accounts—instantly removing livelihoods—without a clear appeal. The opacity also hides wage discrimination and labour-law violations, blocking regulators and unions from spotting unfair practices. CDT says legally mandated data disclosures and accountability are essential before algorithmic management becomes entrenched, requiring effective oversight by public bodies.

Facial recognition, biometric identification, surveillance 

• US: New ICE mobile app pushes biometric policing onto American streets: Biometric Update reports that US Immigration and Custom Enforcement (ICE) has rolled out Mobile Fortify, a smartphone tool that lets agents take a face photo or contact-free fingerprint and receive an automated match against Department of Homeland Security’s IDENT/HART databases in seconds. By shifting biometric verification from fixed checkpoints to city streets, the AI-driven app enables on-the-spot immigration decisions without warrants or human double-checks. Audits reveal lax device security that could leak sensitive data, while civil-liberties groups warn misidentifications and pervasive surveillance threaten Fourth-Amendment protections.
• Hungary: Biometric anti-queer hostility in Hungary: Petition calls on EU Commission to fully ban face recognition: AlgorithmWatch and allies have petitioned the European Commission to ban live facial-recognition after Hungary authorised police to use it against Budapest Pride. The group argues biometric mass-surveillance would let autonomous ID systems flag and track demonstrators, criminalising assembly with no judicial oversight. Policy chief Kilian Vieth-Ditlmann says Hungary’s move could trigger “a snowball effect” of biometric repression unless the EU enforces AI-Act safeguards that already prohibit such deployments.
• Ireland: DPC announces conclusion of investigation into use of facial matching technology in connection with the Public Services Card by the Department of Social Protection: The Data Protection Commission’s final decision finds Ireland’s Department of Social Protection lacked a lawful basis to collect, retain and algorithmically match facial templates for “SAFE 2” Public Services Card enrolment covering 70 % of residents. It imposed a €550,000 fine and ordered processing to stop within nine months unless new legislation sets explicit safeguards. Commissioners warn that large-scale, automated ID checks without clear, transparent rules risk “arbitrary interferences” with citizens’ fundamental rights and data.
• US: Internal struggle over US federal biometric data management intensifies: Biometric Update reports that a internal White House plan would shift DHS’s Office of Biometric Identity Management—and its 320 million-record IDENT/HART databases—under Customs and Border Protection. Critics say giving the nation’s largest biometric hub to a frontline enforcement agency would centralise identity decisions, weaken inter-agency checks and widen autonomous surveillance from immigration to benefits, voting and travel. Sources warn DOGE (Department of Government Efficiency) technocrats could manipulate datasets during the transition, causing system failures, unlawful targeting and loss of privacy safeguards embedded in Office of Biometric Identity Management’s neutral mandate.