News briefing 19 March – 1 April 2025

Autonomy in weapons systems, and AI in security and defence

• India: New arms race: India must help regulate autonomous weapons: Pooja Arora from Jawaharlal Nehru University writes for the Sunday Guardian about the threat of AWS.  According to Arora ‘handover of lethal decision authority represents a profound change in warfare’ and highlights the urgent need for new legally binding rules. AWS goes beyond lowering the threshold for war, but also delegates decisions of life and death to a machine. The author argues that by removing a ‘conscious moral agent deliberating on proportionality and trying to minimize unnecessary suffering’, we risk further dehumanising war. 
• Uruguay: Artificial Intelligence in Security and Defense: Ally or Risk for Uruguay? This interesting article written by a retired Colonel of the Uruguayan Army Pedro M. Gómez explores the potential applications and challenges associated with the use of AI within Uruguay’s security and defense sectors. AI as we know has the capability to improve crime prevention, enhance military efficiency, and strengthen cybersecurity. However this also presents ‘risks such as privacy violations and bias’. Although Uruguay has initiated AI programs, there is a lack of independent oversight. Gómez advocates for stronger ethical regulations, increased human oversight, and expanded AI training for security.
• Ukraine: Military AI: Lessons from Ukraine: Ukraine’s military use of AI is evolving, focusing on enhancing human decision-making and is said to be ‘reducing soldier exposure’. Globally it is known that AI development began after the 2014 annexation of Crimea and was formalized in 2022 to meet needs from the ongoing conflict with Russia. Challenges remain, such as a lack of a clear long-term strategy and ethical concerns. Establishing international norms and legal frameworks will be crucial as AI continues to shape global military strategies and policies.
• International: Move fast, kill things: the tech startups trying to reinvent defence with Silicon Valley values: Tech startups continue to shape  the global defense industry with innovations in AI, drones, and autonomous systems. Companies such as Skydio, Anduril Industries, and Epirus are developing technologies such as autonomous drones, AI pilots, and counter-drone systems and therefore are challenging traditional defense giants. Their rapid development methods, supported by significant venture capital, aim to deliver effective and economical solutions for modern warfare. However, concerns about the reliability and ethical implications of these unproven technologies necessitate thorough testing and comprehensive regulations. 
• Belgium: Commission rejects Groen proposal to ban ‘killer robots’: The Brussels Times reports that ‘a bill proposed by the Flemish Greens to ban fully autonomous lethal weapons systems was rejected by the Chamber’s Defence Commission.’ Members of the opposition parties expressed intent to vote against the bill as it would ‘hinder research and development in the defence industry.’ 

AI, algorithms and autonomy in the wider world

• International: What are AI hallucinations? Why AIs sometimes make things up: This piece in The Conversation on AI hallucinations is by Anna Choi, Ph.D. Candidate in Information Science, Cornell University and Katelyn Xiaoying Mei, Ph.D. Student in Information Science, University of Washington. They refer to AI hallucinations as when an AI system generates ‘plausible but actually inaccurate or misleading’ content. The authors suggest caution and vigilance while using generative AI. 
• International: Paris Summit Showed That Policy Makers Ignore How AI is Harming People Everyday: In this piece published by the Wire, Erika Guevara Rosas critiques the Paris AI summit, accusing policymakers of ignoring the everyday harm caused by AI technologies. She argues that discussions focused more on economic competition and AI sovereignty than on addressing human rights concerns, such as AI’s role in amplifying racial bias in policing, social services, and warfare. Rosas further highlights how predictive policing and facial recognition systems are harming marginalized communities and how automated decision-making in welfare systems is depriving people of basic needs. She warns that the upcoming AI summit in India is likely to overlook these issues, given the country’s history of AI misuse.

Facial recognition, biometric identification, surveillance

• Scotland: Scottish police hold almost no data on facial recognition use: Scotland’s retrospective facial recognition technology is under review again due to insufficient data and oversight. ADR has captured stories relevant to this and between April 2023 and March 2024, ‘Police Scotland performed 3,813 RFR searches, with only 2% matches’.its apparent that this is an issue of poor-quality custody images and no clear biometric technology strategy. Recommendations have included ‘enhancing data collection, evaluations, image quality, clearer policies, better training, and alignment with national initiatives like the UK’s Strategic Facial Matching Project. Increased transparency, governance, and public trust in law enforcement’s use of facial recognition technology are needed.’
• China: China strengthening face biometrics regulation to mandate choice, consent: A new regulation known as ‘Administrative Measures on the Security of Facial Recognition Technology Application’ has been published by the Cyberspace Administration of China (CAC), as reported by Biometric Update. The regulation includes provisions for ‘explicit and informed’ consent before processing a person’s biometric data. 
• UK: UK Home Office, police body resist biometrics transparency: Biometric Update reports on the lack of transparency in the use of facial recognition technologies by the UK Home Office at airports, and by the UK Police. 
• International: Corsight first in facial recognition certified to ISO AI explainability standard: Biometric Update reports that Corsight AI has received certification for compliance with the new ‘ISO/IEC 42001:2023 standard’, which governs AI governance, transparency, and accountability. The standard, effective from 2025, ‘emphasizes AI explainability’, ensuring users understand how AI algorithms determine identity matches or mismatches. ADR recently covered a story about Corsight that brought attention to concerns around surveillance and use in Gaza. 
• Brazil: Brazil regulator demands details on stadium biometrics: Brazil’s National Data Protection Authority (ANPD) is requiring 23 football clubs that use facial recognition for ticketing and stadium access to submit Data Protection Impact Assessments (DPIA). This follows concerns over compliance with Brazil’s General Data Protection Law (LGPD), particularly regarding transparency and the handling of biometric data, including that of fans under 16.