University of Michigan

The University of Michigan (U-M) was founded in 1817 in Detroit as the Catholepistemiad, making it Michigan's oldest university. It relocated to Ann Arbor in 1837 and has grown into one of the nation's preeminent public research universities, with a faculty spanning virtually all academic disciplines. In the context of AI safety, U-M is tracked as a recipient institution for grants supporting faculty research on alignment and oversight problems. Professor Lu Wang leads the LAUNCH Lab in the Computer Science and Engineering (CSE) department. Her work focuses on building trustworthy language models that produce factual, accurate, and safe content. Her most directly AI-safety-relevant project, funded by Open Philanthropy, investigates out-of-context learning in alignment faking — studying how and whether AI models might conceal their true capabilities or behaviors during evaluation. Her research has also been funded by NSF (including a CAREER award), AFOSR, IARPA, and industry partners including LG AI Research and Cisco Systems. Professor Samet Oymak leads the SOTA Lab in the Electrical Engineering and Computer Science (EECS) department. His research focuses on scalable oversight of language model agents, monitoring agentic systems, and the theoretical foundations of transformers and reinforcement learning for language models. His work on VerificAgent demonstrates domain-specific memory verification as a mechanism for scalable oversight of computer-use agents. He has received research support from OpenAI, Google, NSF, Amazon, and Adobe. The Michigan AI Safety Initiative (MAISI) is a student-led organization at U-M dedicated to building the AI safety community on campus, running educational programs such as AI Safety Fundamentals and Intro to ML Safety, and supporting students pursuing AI safety careers and research. Anna Edmonds, a lecturer in the Philosophy department, has received Open Philanthropy support for course development in ethical reasoning. U-M also joined OpenAI's NextGenAI consortium in 2025, and has partnered with Los Alamos National Laboratory on a major AI and high-performance computing facility.