Yildirim on On Artificial Intelligence and Network Effects

Pinar Yildirim (U Pennsylvania The Wharton) has posted “On Artificial Intelligence and Network Effects” on SSRN. Here is the abstract:

Network effects have long been identified as a significant driver of growth for digital platforms. Emergence of artificial intelligence (AI) technologies stands to interact with network effects in significant ways. While several scholars argued that network effects can accelerate the success of AI, it remains less clear how AI-enabled tools themselves might reshape the competitive advantage digital platforms gain from network effects. In this article, I examine the implications of AI tools for network effects. I argue that while some use cases of AI can amplify network effects, others may weaken them. In particular, when the AI tools reduce search and production costs and reduces shared experiences among consumers, AI may reduce the importance of network effects to a digital platform. The paper concludes with the note that new technologies such as AI can have important implications for competition policy and antitrust enforcement.

Trout on When Does Regulation by Insurance Work? The Case of Frontier AI

Cristian Trout (Artificial Intelligence Underwriting Company) has posted “When Does Regulation by Insurance Work? The Case of Frontier AI” on SSRN. Here is the abstract:

No one doubts the utility of insurance for its ability to spread risk or streamline claims management; much debated is when and how insurance uptake can improve welfare byreducing harm, despite moral hazard. Proponents and dissenters of “regulation by insurance” have now documented a number of cases of insurers succeeding or failing to have such anet regulatory effect (in contrast with a net hazard effect). Collecting these examples together and drawing on an extensive economics literature, this Article develops a principled framework for evaluating insurance uptake’s effect in a given context. The presence of certain distortions – including judgment-proofness, competitive dynamics, and behavioral biases – createspotential for a net regulatory effect. How much of that potential gets realized then depends on the type of policyholder, type of risk, type of insurer, and the structure of the insurance market. The analysis suggests regulation by insurance can be particularly effective for catastrophic non-product accidents where market mechanisms provide insufficient discipline and psychological biases are strongest. As a demonstration, the framework is applied to the frontier AI industry, revealing significant potential for a net regulatory effect but also the need for policy intervention to realize that potential. One option is a carefully designed mandate that encourages forming a specialized insurer or mutual, focuses on catastrophic rather than routine risks, and bars pure captives.

Thudumu on How to Measure ROI for AI

Srikanth Thudumu (Institute Applied Artificial Intelligence and Robotics (IAAIR)) has posted “How to Measure ROI for AI” on SSRN. Here is the abstract:

Return on Investment (ROI) is often used as the primary metric for evaluating Artificial Intelligence (AI) projects. However, conventional ROI calculations tend to focus narrowly on short-term, directly attributable savings while overlooking enabling capabilities, strategic options, and risk reduction. Historical technology shifts such as the automobile, electrification, and the internet reveal that value typically emerges after complementary investments and operational redesign. This working paper explains why conventional ROI lenses can be misleading, distills lessons from past transformations, and proposes a simple “Smart ROI” framework with a practical measurement playbook for organizations.

Smith et al. on Regulating Robotaxis

Bryant Walker Smith (U South Carolina Joseph F. Rice Law) and Matthew Wansley (Yeshiva U Benjamin N. Cardozo Law) have posted “Regulating Robotaxis” (99 Southern California Law Review 603 (2026)) on SSRN. Here is the abstract:

In several sunbelt cities, commercial robotaxi service has arrived. The leading robotaxi company is providing over 400,000 trips per week. The industry claims that robotaxis will save lives and provide convenient and affordable mobility. Critics counter that they will increase congestion, undermine transit, and subject the public to ubiquitous surveillance. We argue that the social impact of robotaxis depends on how they are regulated. We emphasize two points missing from the debate. First, some of the benefits of robotaxis may be political rather than technological—some longstanding public policy goals may become viable in a robotaxi world. Second, letting one private company dominate the transportation system risks monopoly abuse—and regulators can act now to prevent it.

In this Article, we offer a plan to regulate robotaxis. Carefully crafted externality regulation can address pollution, congestion, wear-and-tear on infrastructure, and privacy risks while minimizing distortions in choices between travel modes. Regulators can promote competition by permitting open entry, banning lock-in contracts, and enabling one-stop access to competing networks. And they can protect riders even if competition fails by mandating that fares be transparent and rider-neutral and requiring that robotaxi companies maintain a fleet sufficient for emergencies. Policymakers should take advantage of robotaxi deployment to reimagine the transportation system—liberate land from the tyranny of parking, refocus mass transit investments on high-throughput routes, and expand mobility for people with low incomes and people with disabilities.

Remolina on Agentic Payments: When is a Payment (Un)Authorised?

Nydia Remolina (Singapore Management U Yong Pung How Law) has posted “Agentic Payments: When is a Payment (Un)Authorised?” on SSRN. Here is the abstract:

What if your wallet could decide how to spend your money? Payments are no longer made by people alone, they are increasingly executed by software acting on their behalf. These AI agents operate within user-defined parameters, but they also optimise, adapt, and, at times, act in ways the user did not specifically anticipate. This shift challenges the core assumptions of payment law, which remains anchored in binary distinctions between authorised and unauthorised transactions and in models of human intent. Current regulations like PSD2, and the proposed PSR and PSD3, struggle to define when a broad user instruction constitutes valid permission for a specific payment. To bridge this gap, the paper proposes a “bounded delegated authorisation” test, which requires that an agent’s actions stay within strictly defined, provable constraints. The paper also suggests creating a new regulatory category for Digital Assistant Payment Services (DAPS) to ensure accountability and user control. Ultimately, the article argues for a framework where payment service providers must reimburse users if an AI agent initiates a transaction that exceeds its bounded delegated authorisation.

Lajovic on Foundation Models as Infringers: Should Large-Scale AI Training Trigger Collective Licensing Obligations under EU Law?

Tomaž Lajovic (Splato) has posted “Foundation Models as Infringers: Should Large-Scale AI Training Trigger Collective Licensing Obligations under EU Law?” on SSRN. Here is the abstract:

The essay addresses the copyright implications of foundation models (FMs), particularly large language models, under EU law. It examines how training datasets, memorisation, and output generation implicate copyright and database rights, and assesses whether collective licensing regimes or exemptions with remuneration rights could balance the interests of AI developers and rightsholders.

Hartzog et al. on Against AI Half Measures

Woodrow Hartzog (Boston U Law) et al. have posted “Against AI Half Measures” (77 Florida Law Review (forthcoming 2026)) on SSRN. Here is the abstract:

So far, U.S. policy for artificial intelligence has largely consisted of industry-led approaches like encouraging transparency, mitigating bias, promoting principles of ethics, and empowering people. These approaches are vital, but they are only half measures. To bring AI within the rule of law, lawmakers must start drawing substantive lines. 

In this essay, we identify four AI regulatory approaches as half measures. First, transparency does not produce accountability on its own. Next, while mitigating bias in AI systems is critical, even unbiased systems are a threat to the vulnerable. Third, while “AI ethics” are important, they are a poor substitute for laws. Finally, empowering people in their individual choices misses the larger questions about the distribution of power and collective wellbeing.

Instead of these half measures, we recommend that lawmakers reject the idea that AI systems are neutral and inevitable. When lawmakers go straight to putting up half-hearted guardrails, they fail to ask the existential question about whether some AI systems should exist at all. To avoid the mistakes of the past, lawmakers must make the hard calls. And AI half measures will certainly not be enough.

Eto et al. on Beyond Belmont: Convening a National Initiative to Update U.S. Research Ethics Principles for the Age of AI

Tamiko Eto (Stanford U) and Heather Miller (Independent Consultant) have posted “Beyond Belmont: Convening a National Initiative to Update U.S. Research Ethics Principles for the Age of AI” on SSRN. Here is the abstract:

The Belmont Report’s principles have defined U.S. research ethics for decades. However, the rapid evolution of data-driven and AI research has exposed critical gaps that these principles cannot address (Lin & Zhicheng, 2024). While the foundational values outlined in the Belmont Report (respect for persons, beneficence, and justice) remain, they do not guide today’s research ethics or AI research risks, especially with regard to “human data subjects” whose data powers AI systems even when they are not traditional research participants. This paper presents empirical research into IRB professional experiences with AI research oversight and a proposed national initiative to address oversight gaps. Through qualitative case study interviews with Institutional Review Board (IRB)/Human Research Protection Program (HRPP) professionals, we document the specific challenges these regulatory bodies face when reviewing AI research protocols, their familiarity with the application of existing ethical frameworks, and their perspectives on updating foundational principles. Data collection is currently underway. Preliminary findings will ground our proposal for a public-engagement initiative, provisionally termed “Belmont 2.0,” which we are convening with support from the Center for AI & Digital Policy (CAIDP). Full analysis will be presented at the International Association for Safe & Ethical AI (IASEA) conference in February 2026. We address this urgent need through a coordinated, multi-stakeholder effort to update U.S. research ethics principles to meet the unprecedented challenges presented by the digital age. We are seeking organizational partners to co-lead this work. What emerges could become a defining ethical foundation for 21st-century research globally, not only in the United States.

Chen et al. on The “Zhejiang Model” from China: A Novel Framework for AI Ethics Governance Through Leading-Enterprise-Driven Symbiosis

Ye Chen (affiliation not provided to SSRN) et al. have posted “The “Zhejiang Model” from China: A Novel Framework for AI Ethics Governance Through Leading-Enterprise-Driven Symbiosis” on SSRN. Here is the abstract:

The rapid advancement of artificial intelligence (AI) introduces not only socio-economic benefits but also significant ethical risks, including algorithmic bias, data misuse, and loss of control. Addressing these challenges requires innovative governance frameworks that can balance regulation with innovation. This study explores a novel approach from China’s digital economy pioneer, Zhejiang Province, where the collaboration between leading enterprises—categorized as the “Six Dragons of Hangzhou” (e.g., DeepSeek, Unitree Robotics) and the “Six Giants of Zhejiang” (e.g., Alibaba Cloud, Hikvision)—has catalyzed the emergence of a distinctive leading-enterprise-driven symbiosis. We term the resulting systematic framework the “Zhejiang Model.” This model integrates three synergistic components to form a comprehensive governance ecosystem: (a) Two Institutional Innovations: The Zhejiang AI Ethics Governance Regulations and Developer Ethics Liability Insurance establish a “hard constraints + soft safeguards” mechanism. (b) Three Technical Pathways: The AI Ethics & Safety Lab, Industrial Brain Ethics Middleware, and Zhejiang AI Ethics Toolkit operationalize “compliance by design.” (c) A Collaborative Initiative: The Zhejiang AI Ethics Symbiosis Initiative fosters multi-stakeholder co-governance. By articulating this tripartite framework, the study demonstrates how coordinated self-regulation among leading firms, under government guidance, creates a symbiotic ecosystem for AI ethics. The Zhejiang Model thus offers a replicable blueprint for constructing effective, adaptive, and culturally situated AI governance systems.

Zhai et al. on Pseudo Artificial Intelligence Bias

Xiaoming Zhai (The U Georgia) and Joseph Krajcik (Michigan State U CREATE STEM Institute) have posted “Pseudo Artificial Intelligence Bias” on SSRN. Here is the abstract:

Pseudo artificial intelligence bias (PAIB) is broadly disseminated in the literature, which can result in unnecessary AI fear in society, exacerbate the enduring inequities and disparities in access to and sharing the benefits of AI applications, and waste social capital invested in AI research. This study systematically reviews publications in the literature to present three types of PAIBs identified due to (a) misunderstandings, (b) pseudo mechanical bias, and (c) overexpectations. We discuss the consequences of and solutions to PAIBs, including certifying users for AI applications to mitigate AI fears, providing customized user guidance for AI applications, and developing systematic approaches to monitor bias. We concluded that PAIB, due to misunderstandings, pseudo mechanical bias, and overexpectations of algorithmic predictions, is socially harmful.