In an effort to share ideas beyond the limited time we have during the annual conference, the Community Building Committee asked Graduate Students attending TPRC54 to take notes of paper sessions to be made available to the community. The following is a summary of the paper session entitled “AI and the Law” that was presented on Friday, September 25, 2026 at 9:00am. The session was moderated by Gus Hurwitz, University of Pennsylvania. The following is an edited summary based on the notes provided by Boxi Chen, York University.
Paper 1: Polycentric Governance Gaps in U.S. AI and Privacy Law
Isaac Haizel, University of Illinois at Urbana Champaign; Julia Trevino, University of Illinois at Urbana Champaign; Madelyn Rose Sanfilippo, University of Illinois at Urbana Champaign; Ece Gumusel, Rutgers University
This presentation examined how states are responding to the potential harms of artificial intelligence through their own legislative and regulatory frameworks rather than relying on a single, centralized federal approach. The presenter described this as a form of polycentric governance, in which multiple levels of government develop their own approaches to addressing AI-related challenges.
By comparing proposed bills and enacted laws, the presentation highlighted how widely state approaches can vary, particularly in their definitions of AI and the scope of activities covered. As a result, businesses may face different requirements depending on the state in which they operate, making compliance more difficult. At the same time, individuals may receive different levels of protection depending on where they live.
The presentation raised an important question about how to create a more coherent approach to AI governance. Greater consistency across state frameworks could help reduce the compliance burden for businesses while ensuring that people receive adequate protections from potential AI harms.
Paper 2: The Crooked Brush: Against the Privatisation of Copyright Regulation in the Post-AI Era
Zachary Cooper, Vrije Universiteit Amsterdam/Weizenbaum Institute, Australia
This presentation examined how generative AI has created new challenges for copyright law, with the presenter arguing that AI-related infringement has reached an unprecedented scale. The legal questions extend beyond whether an AI model has copied a protected work. Potential infringement can arise throughout the creative process, from the input of copyrighted material during development and training to the generation of outputs. As creative work moves through multiple tools and platforms, tracing how copyrighted material is used becomes increasingly difficult.
The presenter argued that generative AI does not need copyrighted works for training and that training on protected works should be illegal. However, the presentation also raised concerns about the potential consequences of greater restrictions. More restrictive copyright rules could encourage AI companies to create “walled gardens,” giving platforms greater control over the content and tools available within their systems. This could alter the relationship between platforms and artists while further concentrating market power.
Overall, the presentation highlighted the complexity of applying existing copyright protections in the context of generative AI. It raised important questions about how to protect creators' rights while accounting for the evolving tools, platforms, and power dynamics shaping the creative process.
Paper 3: The High-Risk of ‘Small AI Harms’: A Policy Analysis of International Approaches to AI Governance
Boxi Chen, York University, Canada; Jonathan Obar, York University, Canada; Sachil Singh, York University, Canada
This presentation examined how current AI governance frameworks tend to focus on “high-risk” AI systems, potentially overlooking less visible harms that fall outside predefined categories. The presenter argued for expanding the recognition of these “small” harms, drawing on previous work related to personally identifiable information and the challenges of identifying and defining harm. Creating an exhaustive list of AI-related harms is difficult, particularly because harm does not need to be immediate, severe, or easily visible to have meaningful consequences.
The presentation emphasized that seemingly minor harms can become significant through repetition and accumulation, either for an individual or across a larger population. Examples included surveillance pricing, AI-generated misinformation, and the repurposing of healthcare data. These types of harms may not fit neatly into existing high-risk classifications, suggesting that current frameworks are not broad or flexible enough to capture the full range of AI-related risks.
Rather than abandoning high-risk classifications, the presenter proposed expanding existing approaches to create a more flexible framework that recognizes both high-risk and “small” harms. The presentation highlighted the importance of looking beyond predefined categories and considering how seemingly minor impacts can accumulate and affect individuals and society over time.
Session Reflection from Boxi Chen:
“I enjoyed seeing the different ways AI can create social and policy challenges. Overall, the papers made me think more critically about how difficult it is to regulate AI in a way that protects people while still allowing innovation.”