News briefing 11 August 2025

Editorial: 

Digital dehumanisation is not a metaphor but a mechanism. As the Automated Decision Research explainer defines it, it is the process whereby ‘humans are reduced to data, which is then used to make decisions and/or take actions that negatively affect’ them. In the military domain, this manifests when sensing, identification and engagement cycles compress into machine-paced workflows in which the human becomes observer, confirmer—or bystander. The ethical risk is not only error; it is the erosion of personhood and judgment at scale.

Policymakers across Asia and beyond are trying to keep humanitarian law abreast of this shift. The ICRC’s Asia-Pacific Symposium convened military, legal and humanitarian experts in Bangkok to examine how AI and autonomous weapons (AWS) can comply with IHL—an acknowledgement that normative guardrails must be clarified before automation hardens into practice. Dialogue is necessary, but insufficient unless it yields tangible rules on human control, explainability and traceability.

Meanwhile, capability is accelerating. The U.S. Department of Defense has awarded large ‘frontier AI’ contracts even as experts warn of ‘unknowns’ and seek transparency about how models were vetted before award—an adoption curve that risks normalising opaque systems in high-stakes use. In parallel, the open-source frontier is militarising: RSIS flags China’s ChatBIT as a case where freely modifiable models can be adapted for battlefield intelligence and decision-support, outpacing regional oversight. 

Procurement cycles are also compressing. The U.S. Army’s new ‘launched effects’ solicitation calls for attritable autonomous drones able to ‘detect, identify, locate and report,’ with growth paths to automatic target recognition, swarming, and options for inert or live lethal payloads—deliverable to units in 2026 and, in some cases, within ‘four to six months.’ Speed without commensurate safeguards risks substituting database likenesses for human beings and probabilistic matches for legal determinations—textbook dehumanisation. 

Markets amplify the trend. Queen Mary University of London cautions that venture-capital logics—scale, speed, and ‘prototype’ culture—are migrating into defence, with ethical risks when ‘life-and-death decisions’ are mediated by iterative, investor-driven technologies. Incentives that prize velocity and valuation over verification further distance decision-makers from those affected.

The same dynamics bleed into domestic security. London’s plan to more than double live facial-recognition deployments—scanning millions against watchlists—illustrates how mass surveillance can flatten citizens into data subjects; by contrast, New Zealand’s new biometric code demands necessity, proportionality and transparency, showing that rights-preserving governance is possible.

Keeping people at the centre of war-time decisions is not a slogan; it is the minimum ethical architecture for security in an automated age.

Autonomy in weapons systems, and AI in security and defence

• Thailand: Asia-Pacific Symposium discusses IHL and new technologies of warfare: The ICRC notes that the Asia-Pacific Symposium in Bangkok ‘brought together military, legal and humanitarian experts… to explore how emerging methods of warfare should comply with international humanitarian law,’ with discussions centered on ‘the legal, ethical and humanitarian implications’ of ‘artificial intelligence (AI)’ and ‘autonomous weapons systems (AWS).’ Thailand’s Judge Advocate General urged participants to ‘exchange ideas, build networks and collectively chart directions,’ as sessions highlighted tools like the digital emblem and VR training to buttress protection of civilians. The emphasis: regional dialogue to keep IHL aligned with rapid military innovation.
• International: Beware the Hype: Venture capital for military technology poses ethical risks: Queen Mary University of London’s policy brief by Prof. Elke Schwarz warns that venture capital (VC) is reshaping defence with a Silicon-Valley ethos that ‘needs military startups to scale up fast,’ but this ‘move fast and break things’ model poses ‘strong ethical concerns’ where technology mediates ‘life and death decisions.’ It argues VC’s high-risk, high-reward logic—’about 80% of investments fail’—drives overpromising, rapid prototyping  ‘in live conflicts,’ and sidelines transparency and democratic oversight, trends ‘likely to increase global instability.’ The brief traces a U.S. shift to infuse ‘the Pentagon(…) with the nimble, agile DNA of Silicon Valley,’ easing quick contracts and prioritising ‘prototype warfare,’ while citing sky-high valuations (e.g., Anduril) despite no profits. Recommendations urge resisting hype, recognising second- and third-order effects, and ensuring guardrails so speed, scale and experimentation do not eclipse legal and ethical principles in procurement and use.
• US: Experts worry about transparency, unforeseen risks as DOD forges ahead with new frontier AI projects: Reporting on new Pentagon ‘frontier AI’ contracts ‘each worth up to $200 million,’ DefenseScoop flags expert concerns that CDAO is ‘not responding with full transparency regarding if and how those four powerful models were vetted’ before award. While foundation models are ‘rapidly pushing the boundaries,’ the piece cites warnings about ‘unknowns’ that ‘create risks to humanity.’ The core policy tension—accelerated adoption versus safety assurance—turns on how DoD will document evaluations and mitigations for high-stakes operational use.
• US: High-Risk AI Models Need Military-Grade Security: Arguing that ‘high-risk AI models need military-grade security,’ the commentary for War on the Rocks says America’s ‘light-touch regulatory model’ leaves advanced research ‘critically vulnerable to theft and sabotage,’ with Chinese services ‘very likely targeting American AI labs.’ Leading labs are described as ‘the security equivalent of swiss cheese,’ prompting a call to treat them as ‘strategic national assets’ with secure, government-audited data centers, classification regimes for model weights, and rigorous personnel vetting. A ‘tiered risk governance framework’ is urged to match controls to capability and dual-use potential. 
• International/Asia: ChatBIT and the Militarisation of Open-Source AI: Security Implications for Asia: S. Rajaratnam School of International Studies (RSIS) warns ‘the rapid advancement of open-source AI has outpaced regulatory oversight,’ spotlighting China’s ChatBIT, ‘designed to support military operations, including battlefield intelligence, situational awareness, and operational decision-making.’ Because open-source models ‘can be freely modified,’ enforcement of civilian-only acceptable-use policies is ‘a challenge.’ With ASEAN only recently issuing its ‘first joint statement on military AI,’ RSIS urges ‘stronger regional coordination’ and early-warning and transparency mechanisms to address risks that open-source militarisation poses to Asian stability. 
• International: Rising Demand for AI in Military Applications: Bloomberg highlights the ‘Rising Demand for AI in Military Applications,’ noting ‘Palantir shares have jumped more than 500%,’ with investors focusing on ‘its growing presence in AI driven defense.’ The segment situates defense-AI demand amid surging capital expenditure and procurement interest, framing dual-use software and data integration as core drivers. The takeaway is that financial markets are reinforcing defense-sector AI acceleration—an external pressure that complicates efforts to pace deployments with guardrails and transparency.
• US: Army issues solicitation for ‘launched effects’ autonomous drones: On the U.S. Army’s solicitation for ‘launched effects,’ DefenseScoop reports a 2026 push to field autonomous, ‘attritable’ drones able to ‘detect, identify, locate and report’ and ‘autonomously execute assigned missions without requiring direct intervention.’ Desired features include ‘a growth path to automatic target recognition,’ ‘autonomous swarming capabilities,’ and options to ‘operate inert lethal payloads… [and] live lethal payloads.’ The Army seeks interoperable, modular systems deliverable ‘within four to six months,’ signaling rapid acquisition cycles for human-machine teaming at scale.
• International: Governing Lethal Autonomous Weapons in a New Era of Military AI: This piece in TRENDS Research and Analysis argues lethal autonomous weapons systems (LAWS) are ‘rapidly approaching operational use,’ defined as systems that ‘select and engage targets without human intervention.’ It warns that autonomy shifts ‘critical decision-making’ from humans to machines, putting ‘responsibility, proportionality, and distinction’ at risk. With states split between bans, regulation, or status quo, the piece concludes ‘the question is no longer if LAWS should be governed, but how.’ It frames ‘meaningful human control’ and definitional clarity as prerequisites for accountability and humanitarian compliance. 

AI, algorithms and autonomy in the wider world

• UK: Staff at UK’s top AI institute complain to watchdog about its internal culture: The Guardian reports that staff at the Alan Turing Institute filed a whistleblowing complaint alleging governance failures and warning the institute is ‘in danger of collapse.’ The technology secretary reportedly linked future support to ‘improved delivery and leadership reform,’ with a shift toward ‘defence and national security.’ Critics say projects on bias and rights are being shuttered amid ‘lack of accountability.’
• International: What will the AI revolution mean for the global south?: Krystal Maughan for The Guardian cautions that ‘we must avoid inequalities… being perpetuated in the digital age,’ noting promises of AI ‘democratisation’ still ‘benefit mainly countries with access to tech hubs.’ The commentary calls for Global South collaboration to build markets and ‘a model of sovereignty for their data and data labour.’ Inclusion, visa access for researchers, and equitable infrastructure are highlighted as conditions for meaningful participation in AI’s future.
• EU: EU begins rollout of new AI rules with tech giants split on compliance: Reporting on the EU AI Act’s rollout, RFI says ‘the compliance period begins this weekend,’ signalling Brussels will ‘press ahead with its approach to AI governance.’ It describes the start of obligations for general-purpose AI and the push for transparency and safety while industry reactions diverge. The piece underscores that timelines are firm as the EU seeks to operationalize oversight and align providers with new standards amid global regulatory fragmentation. 
• International: Big tech has spent $155bn on AI this year. It’s about to spend hundreds of billions more: The report by The Guardian finds ‘big tech has spent $155bn on AI this year,’ with capital outlays set to soar as Microsoft, Meta, Amazon and Alphabet escalate data-center and chip investments. Microsoft plans to spend ‘about $100bn,’ Meta $66bn to $72bn,’ and Alphabet ‘$85bn,’ taking combined capex ‘to more than $400bn’ next year. The article frames capex as a proxy for AI build-out, noting the infrastructure demands of ‘servers and data centers to support AI.’

Facial recognition, biometric identification, surveillance

• US: Senate Republicans not backing down in battle with TSA over facial recognition bill: POLITICO reports that the senate GOP ‘privacy hawks’ are ‘not backing down’ after a bill to limit airport face-scanning stalled, blaming TSA for ‘orchestrating a lobbying effort.’ The measure would ‘notify every passenger’ of opt-out rights and constrain biometric data retention. Sen. Jeff Merkley likens expanded airport facial recognition to tools ‘used to intimidate… freedom of expression and freedom of assembly.’ The clash pits administrative pushes for automation against legislators seeking explicit civil-liberties guardrails.
• UK: London police can use live facial recognition 10 times a week under new rules: This piece in Biometric Update reports on London’s Metropolitan Police will ‘increase up to five days a week, delivering up to 10 deployments a week’ of live facial recognition, amid staffing cuts and rising crime. The outlet notes ‘police scanned nearly 4.7 million faces’ in 2024 and cites critics warning police are ‘off the leash’ with ‘no existing legislation’ governing the technology. The Met defends LFR as targeting ‘serious offenders,’ pointing to arrests and charges attributed to deployments. 
• New Zealand: Stronger rules planned for use of biometric data: RNZ reports that New Zealand’s Privacy Commissioner announced a biometric processing code introducing ‘stronger and clearer rules’ for facial recognition, fingerprints and iris scans. Agencies must ensure use is ‘necessary, effective and proportionate,’ add safeguards, and inform the public when data are collected. The code bars ‘particularly intrusive’ uses that infer protected attributes and sets special rules for children; compliance is due by August 2026 following a November commencement.