Infra-Legalities: Global Security Infrastructures, Artificial Intelligence and International Law

由于机器学习（ML）和人工智能（AI）的快速发展，人们越来越多地通过新形式的数据分析来识别和应对恐怖主义风险和威胁。包括社交媒体平台、航空公司和金融机构在内的私营行为体目前正积极与各国和国际组织合作，实施雄心勃勃的数据安全项目，以支持全球反恐努力。联合国安理会（UNSC）呼吁所有国家通过建立观察名单和共享生物识别数据来加强有关恐怖分子嫌疑人的信息交流，利用机器学习提前预测识别“未来的恐怖分子”。社交媒体平台正在使用人工智能来检测在线极端主义内容，并以前所未有的规模监管全球数据流。分析航空业的乘客数据，以识别可疑的“行为模式”，并控制危险旅客的行动。银行通过挖掘金融数据来发现可疑交易和恐怖分子的“关联”。这些变化都在推动新的、影响深远的全球信息基础设施项目。然而，这些变化对国际法如何实施、对全球安全威胁的认识以及对强大行为体的问责的影响仍然不确定。作为全球治理基础的数据基础设施在法律学术领域基本上被忽视了。虽然人工智能带来的潜在问题（歧视和侵犯隐私）变得越来越清晰，但解决方案仍然难以捉摸——尤其是在安全领域，保密是关键，算法的内部工作比平时更加“黑盒化”。监管理论家认为，我们迫切需要“扩大我们的权利话语框架，以涵盖我们的社会技术架构”，以应对人工智能的问责挑战（Yeung 2019）。换句话说，数据基础设施可以为重新构想如何在我们的数字时代重新连接信息和权利提供基础。该项目从其创造的“基础设施空间”重新思考全球安全法，重点关注(1)打击网络恐怖主义和(2)控制“危险”个人的活动。我的假设是，全球安全治理最深远的变化不是用国际法的语言写出来的，也不是通过国家和国际组织的正式权力创造出来的，而是通过新的社会技术基础设施和它们所支持的专业知识来实现的。数据基础设施对于理解如何通过人工智能扩展权利至关重要。我提出了“非法行为”（或数据基础设施的监管效应）的概念，以分析这些转变，并为在算法全球治理时代研究国际法和监管开发一种新方法。基础设施通常被视为一种无形的基础，而强大的行动者在其上行动。它很少被看作是可以创造和塑造知识和治理的东西。借鉴科学与技术研究、计算机科学和安全研究，该项目执行了Bowker和Star（1999）所称的“基础设施倒置”，通过映射该领域中看似平凡的数据基础设施治理工作。通过“跟踪数据”——并追踪安全基础设施正在制定的社会技术关系、规范、知识实践和权力不对称——可以出现一种研究全球治理的不同方法。国家、IOs和技术平台都在呼吁人工智能的道德发展。人们提出了不同的监管方法，但对于如何减轻人工智能的不利影响，同时拥抱其巨大的潜力，却没有达成共识。研究全球安全法的违法性，为应对这些挑战和在人工智能和自动化时代塑造当前关于安全、信任和问责制的政策辩论开辟了空间。

Terrorist risks and threats are increasingly identified and countered through new forms of data analytics made possible by rapid advances in machine learning (ML) and artificial intelligence (AI). Private actors, including social media platforms, airlines and financial institutions, now actively collaborate with states and international organisations (IOs) to implement ambitious data-led security projects to support global counterterrorism efforts. The UN Security Council (UNSC) has called on all states to intensify the exchange of information about suspected terrorists by building watchlists and sharing biometric data, using ML to predictively identify 'future terrorists' in advance. Social media platforms are using AI to detect extremist content online and regulate global data flows on an unprecedented scale. Passenger data from the aviation industry is analysed to identify suspicious 'patterns of behaviour' and control the movements of risky travellers. Financial data is mined by banks to spot suspicious transactions and terrorist 'associations'. These changes are all putting new and far-reaching global information infrastructure projects into motion. Yet the implications of these shifts for how international law is practiced, global security threats known and powerful actors held accountable remain uncertain. The data infrastructures underlying global governance have been largely neglected in legal scholarship. And whilst potential problems that AI poses (discrimination and privacy violations) are becoming clearer, solutions remain elusive - especially in the security domain, where secrecy is key and the inner workings of algorithms are 'black-boxed' even more than usual. Regulatory theorists argue that we urgently need to 'expand our frame of rights discourse to encompass our socio-technical architecture' to respond to the accountability challenges of AI (Yeung 2019). Data infrastructures, in other words, might provide the basis for reimagining how information and rights could be reconnected in our digital present. This project rethinks global security law from the 'infrastructure space' it is creating, focusing on (i) countering terrorism online and (ii) controlling the movements of 'risky' individuals. My hypothesis is that the most far-reaching changes to global security governance are not being written in the language of international law, or created through the formal powers of states and IOs, but built through new socio-technical infrastructures and the expertise they are enabling. Data infrastructures are critical for understanding how rights might be extended through AI. I develop the concept of 'infra-legalities' (or, the regulatory effects of data infrastructures) to analyse these shifts and develop a new approach for studying international law and regulation in the age of algorithmic global governance. Infrastructure is usually disregarded as an invisible substrate on which powerful actors act. It is rarely seen as something through which knowledge and governance can be created and shaped. Drawing from Science and Technology Studies, computer science and security studies, this project performs what Bowker and Star (1999) call an 'infrastructural inversion' by mapping the seemingly mundane governance work of data infrastructures in this domain. By 'following the data' - and tracing the socio-technical relations, norms, knowledge practices and power asymmetries that security infrastructures are enacting - a different method of studying global governance can emerge. States, IOs and tech platforms are all calling for the ethical development of AI. Different regulatory approaches are proposed with no consensus on how to mitigate the adverse effects of AI whilst embracing its vast potentialities. Studying the infra-legalities of global security law opens space for addressing these challenges and shaping current policy debates on security, trust and accountability in the age of AI and automation.