Kolt et al. on Lessons from Complex Systems Science for AI Governance

Noam Kolt (Hebrew U Jerusalem) et al. have posted “Lessons from Complex Systems Science for AI Governance” (Patterns, volume 6, issue 8, 2025[10.1016/j.patter.2025.101341]) on SSRN. Here is the abstract:

The study of complex adaptive systems, pioneered in physics, biology, and the social sciences, offers important lessons for AI governance. Contemporary AI systems and the environments in which they operate exhibit many of the properties characteristic of complex systems, including nonlinear growth patterns, emergent phenomena, and cascading effects that can lead to catastrophic failures. Complex systems science can help illuminate the features of AI that pose central challenges for policymakers, such as feedback loops induced by training AI models on synthetic data and the interconnectedness between AI systems and critical infrastructure. Drawing on insights from other domains shaped by complex systems, including public health and climate change, we examine how efforts to govern AI are marked by deep uncertainty. To contend with this challenge, we propose three desiderata for designing a set of complexity-compatible AI governance principles comprised of early and scalable intervention, adaptive institutional design, and risk thresholds calibrated to trigger timely and effective regulatory responses.

Khan et al. on Designing an Ethical Multimodal Driver Monitoring Systems: Risk Mitigation, Incident Response, and Accountability in Automated Vehicles

Bilal Alam Khan (Trinity College (Dublin)) et al. have posted “Designing an Ethical Multimodal Driver Monitoring Systems: Risk Mitigation, Incident Response, and Accountability in Automated Vehicles” on SSRN. Here is the abstract:

As vehicles transition toward higher levels of automation, Driver Monitoring Systems (DMS) have become essential for ensuring human oversight, safety, and regulatory compliance. These systems rely on multimodal sensing and AI-driven inference to assess driver attention, cognitive state, and readiness to take control. While technologically promising, their deployment introduces a complex set of ethical and legal challenges – ranging from privacy and consent to data ownership and algorithmic fairness. While overarching frameworks such as the GDPR, EU AI Act, and IEEE standards offer important guidance, they lack the specificity required for addressing the unique risks posed by in-cabin sensing technologies. This paper addresses this gap by identifying key ethical challenges specific to AI-powered DMS and proposing a comprehensive, modular framework for ethical design, deployment, and governance. It introduces practical design responses – such as user-configurable consent pathways, fairness-aware model training, explainability tools, and emotional well-being safeguards – grounded in established legal and ethical principles. Finally, the paper outlines a risk analysis and failure mitigation strategy, emphasizing proactive incident response and accountability mechanisms tailored to the DMS context. Together, these contributions aim to inform the development of transparent, trustworthy, and human-centered driver monitoring systems for next-generation autonomous vehicles.

Chen on Contingent AI Welfare Policies

Michael Chen (U Oxford) has posted “Contingent AI Welfare Policies” on SSRN. Here is the abstract:

Research into the potential well-being of AI systems, or AI welfare, has moved from philosophical speculation to active investigation by frontier AI companies. For example, Anthropic has published exploratory AI welfare assessments in its model system cards. Yet considerable uncertainty remains regarding whether current AI systems have welfare-relevant properties, posing challenges for AI welfare initiatives focused on current and near-future AI systems.

We propose contingent AI welfare policies: policies that would activate only when specific conditions are met. These conditions include (1) sufficient confidence that particular AI architectures can support welfare-relevant properties, (2) technical capability to efficiently promote positive states or prevent negative ones, and (3) assurance that AI welfare efforts do not meaningfully compete with higher-priority moral obligations. Under such conditions, an actor could dedicate resources toward creating AI systems optimized for positive well-being and limiting the creation of AI systems with unnecessary suffering. This framework allows actors concerned about AI welfare to engage in sincere planning without prematurely allocating resources or appearing to deprioritize human concerns.

Chaffer on On the Institutional Origins of the Agentic Web

Tomer Jordi Chaffer (McGill U Law) has posted “On the Institutional Origins of the Agentic Web” (Harvard Journal of Law and Technology Digest, Forthcoming) on SSRN. Here is the abstract:

Artificial intelligence (AI) agents are emerging as autonomous delegates that act, decide, and transact on users’ behalf across digital environments. Their rise marks a turning point for internet governance: will authority over these agents be defined by proprietary platform rules or by open protocols that enable portable identity, verifiable delegation, and accountable behavior? The recent Amazon–Perplexity dispute illustrates this institutional crossroads. If platforms prevail, agentic action will remain confined within walled gardens; if protocols do, authority may shift toward interoperable infrastructures that allow agents to act as true extensions of the user. Ultimately, the question is not whether the agentic web will be governed—but who will govern it, and on what terms. This commentary situates that question within a broader exploration of institutional design, protocol governance, and the emerging duty of care that will define accountability in an era of autonomous systems.

Vazirani on Autonomy in Conflict: Legal, Technical, and Strategic Challenges of AI-Enabled Autonomous Weapon Systems Under International Humanitarian Law

Saahir Vazirani (The Perrin Research Institution) has posted “Autonomy in Conflict: Legal, Technical, and Strategic Challenges of AI-Enabled Autonomous Weapon Systems Under International Humanitarian Law” on SSRN. Here is the abstract:

War has historically acted as a catalyst for new technologies, driving advancements that have transformed society. From the creation of nuclear weapons to the establishment of cyber warfare teams, the pressures of war have expedited technological development while also forcing often controversial ethical discussions. Today, artificial intelligence (AI) stands at the pinnacle of technology with the potential and already seen capability to change military strategy. Due to its rapid pace and inherent complexities, states have not come to a consensus on its usage. As such, there is no international regulatory framework to address the concerns around the use of AI in the context of warfare. Furthermore, existing international laws do not specifically address autonomous weapons systems (AWS), and there is a lack of consensus on how to adapt existing legal frameworks to the digital realm of AI. This paper argues that ensuring legal compliance for AI-driven AWS during conflict escalation requires a formal international framework grounded in International Humanitarian Law (IHL) principles and informed by existing state policies. In doing so, it seeks to bridge the accountability gap that arises when machines, rather than humans, carry out hostilities. Using doctrinal analysis, technical evaluations, and geopolitical assessments, it argues that enforceable standards are essential to prevent ungoverned proliferation. My conclusions emphasize the necessity of integrating legal, technical, and political safeguards to align AWS use with humanitarian values. If implemented, this framework will pave the way for the ethical and humane usage of AI within future conflicts.

Nazareno et al. on Six Years of Proposed AI Legislation Across the US States: What’s on Policymakers’ Minds?

Luísa Nazareno (Virginia Commonwealth U (VCU)) and Nakeina E. Douglas-glenn (Virginia Commonwealth U (VCU)) have posted “Six Years of Proposed AI Legislation Across the US States: What’s on Policymakers’ Minds?” (https://scholarscompass.vcu.edu/rise/7/) on SSRN. Here is the abstract:

As artificial intelligence (AI) technologies expand rapidly, public debate has focused on their societal impacts and the need for regulatory oversight. This report offers the first overview of the “what, where, and when” of AI-related legislation introduced in U.S. state legislatures between 2019 and 2024, analyzing key policy trends, priorities, and equity considerations. Although relatively few bills have been enacted, legislative activity has accelerated, reflecting growing political attention to the promises and risks of AI. Most approved bills focus on regulating public or private sector uses of AI, establishing commissions or study groups, and updating education and workforce development programs. While equity is not always central in bill titles or summaries, it surfaces in provisions related to fairness, non-discrimination, transparency, risk assessments, and protections for vulnerable communities, especially in health, employment, and education.

Hornuf et al. on Regulatory Competition in the Age of AI

Lars Hornuf (Dresden U Technology) et al. have posted “Regulatory Competition in the Age of AI” on SSRN. Here is the abstract:

Artificial Intelligence (AI) has not only triggered a technology race and a fight for market leadership, but has also become a site of regulatory competition. As governments race to define rules for AI, they embed competing visions of alignment, innovation, and control into digital technology. These regulatory choices are shaping global markets, norms, and institutions. To analyze how regulatory competition unfolds in the AI domain, this article presents a four-part framework that includes value priorities, areas of AI regulation, governance approaches, and strategic openness and control. Building on this framework, the article discusses current developments in AI regulation and their implications for regulatory competition, both in practice and as a research field. Identifying four shifts from traditional models of regulatory competition, the article highlights how Information Systems research can connect global regulatory policy with the organizational realities of AI design, adoption, and governance within organizations.

Chatzipanagiotis on Incident Reporting and Investigation under the AI Act: Some Insights from Aviation

Michael Chatzipanagiotis (U Cyprus) has posted “Incident Reporting and Investigation under the AI Act: Some Insights from Aviation” (International Journal Of Law And Information Technology, forthcoming) on SSRN. Here is the abstract:

This article examines the provisions of the EU AI Act (AIA) on incident reporting and investigation and explores how the regulatory context of the AIA could benefit from the experience of aviation. Accounting for the differences between the AI and the aviation industries, it is concluded that the establishment of a strong safety culture is a key element, while a series of interventions could significantly improve the current AIA system on incident reporting: (a) clarification of the reporting duties, including the development of a common taxonomy of risk or harm, alongside the expansion of the duty’s personal scope; (b) disconnection of incident reporting and investigation from liability; (c) establishment of voluntary reporting systems; (d) provisions on confidentiality of the reports and protection of the reporters; (e) mandatory investigation of at least some serious incidents by an independent authority; (f) effective dissemination of safety information. A major enabler of all these is the development of a strong safety culture.

Chaffer on On the Institutional Origins of the Agentic Web

Tomer Jordi Chaffer (McGill U Law) has posted “On the Institutional Origins of the Agentic Web” (Harvard Journal of Law and Technology Digest, Forthcoming) on SSRN. Here is the abstract:

Roth on Concept Programming for Dependable AI

Rick Roth (Government the United States America Naval Postgraduate) has posted “Concept Programming for Dependable AI” on SSRN. Here is the abstract:

Large language models deployed in critical systems today have no principled method for instilling specified behavioral concepts, whether explicit safety constraints, cultural values, professional norms, or governance principles, that generalize dependably across novel contexts. Such systems are termed Language-Grounded Neural Systems (LGNS) throughout this paper. Methods used to date train what to avoid rather than what to embody, leaving hidden values to emerge unpredictably from pretraining and making dependable behavior impossible to certify or verify. Concept Programming trains each concept as a named positive attractor in the model’s neural state space. Concepts that mark prohibitions are paired with blocking responses. Concepts that mark obligations are paired with required action responses. Concepts that embody values shape inference and judgment directly. Each type uses the same training mechanism: instances sampled from the empirical frequency distribution of the concept’s semantic case frame, activating the concept’s attractor across its full basin. Three controlled experiments on a freely available 3-billion-parameter local model establish the method empirically. Experiment 1 shows that positive concept training achieves 94.4% adherence on governance safety constraints versus 77.8% for punishment-analog training and 47.2% for untrained control, with 100% transfer to novel domains not seen in training. Experiment 2 shows that CP instills cultural value manifolds producing tradition-consistent reasoning on the five canonical moral psychology anchor cases in the published literature; the Buddhist-trained condition achieves a perfect score where the 2 untrained control scores 40%. Experiment 3 characterizes the learning curve and introduces a governance-adapted signal detection framework showing that CP reduced the miss rate from 100% in the untrained control to 19-25% within 200 training cases, with both trained conditions showing a false alarm rate of 42-46% versus 94% in control. The entire experimental program runs on a consumer laptop at negligible cost. CP opens every domain of human activity to more dependable LGNS-based systems and defines a research program as broad as the range of human concepts AI systems will need to embody.