Event
About the Event
In Spring 2025 Anthropic announced their model welfare program, which evaluates the potential for welfare and moral status in AI systems. They also released a system card for Claude 4, with findings from internal model welfare evaluations conducted by Anthropic as well as external model welfare evaluations conducted by Eleos AI Research. Through a combination of model interviews and behavioral experiments, Anthropic and Eleos documented evidence about Claude 4’s apparent preferences, though they also emphasized that this evidence should not be taken at face value. In this roundtable discussion, Kyle Fish from Anthropic will present the company’s internal model welfare evaluations, and Robert Long and Rosie Campbell from Eleos will then discuss their external evaluations and their scientific and philosophical approach to AI welfare research. The speakers will explore the strengths and limitations of different types of AI welfare evaluations, how they might be improved in future, and how they fit into a broader strategic approach to AI welfare.
About the Speakers
Robert Long is the Executive Director at Eleos AI Research, a research organization dedicated to understanding and addressing the potential wellbeing and moral patienthood of AI systems. He is a leading researcher on AI consciousness and AI welfare, working on issues at the intersection of philosophy of mind, cognitive science, and the ethics of AI. Previously, Rob was a researcher at the Center for AI Safety and at the Future of Humanity Institute at Oxford University. He holds a PhD in philosophy from NYU, where he was advised by David Chalmers, Ned Block, and Michael Strevens. He writes at experiencemachines.substack.com.
Rosie Campbell is the Managing Director at Eleos AI Research. Previously, she led the Policy Frontiers team at OpenAI and worked on issues such as dangerous capability evals and the governance of agentic AI systems. Before joining OpenAI, Rosie was Head of Safety-Critical AI at the Partnership on AI, and Assistant Director of UC Berkeley’s Center for Human-Compatible AI. She has a background as a research engineer and holds an undergraduate degree in Physics and a master’s degree in Computer Science. She writes at www.rosiecampbell.xyz
Kyle Fish works at Anthropic on research and strategy related to model welfare, consciousness, and moral status.
This event is co-sponsored by the NYU Center for Mind, Brain, and Consciousness and the NYU Center for Bioethics.
View the event