Michigan State University's Department of Computer Science and Engineering (CSE) conducts AI safety research, notably through the OPTML group's work on trustworthy machine learning and LLM unlearning.
Michigan State University's Department of Computer Science and Engineering (CSE) conducts AI safety research, notably through the OPTML group's work on trustworthy machine learning and LLM unlearning.
People
Updated 05/18/26Red Cedar Distinguished Associate Professor, Computer Science and Engineering
Funding Details
Updated 05/18/26- Annual Budget
- $8,000,000
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 05/18/26Michigan State University's Department of Computer Science and Engineering (CSE) is a large, research-active academic department in East Lansing, Michigan. It houses approximately 45 faculty members, over 20 research laboratories, and awards roughly 400 bachelor's, 40 master's, and 20 PhD degrees annually. Annual research expenditures exceed $8 million, supported by federal agencies including NSF, DARPA, ARO, and DOE, as well as corporate sponsors. The department's most prominent AI-safety-relevant research group is OPTML (OPTimization and Trustworthy Machine Learning), led by Sijia Liu. Liu was promoted to Red Cedar Distinguished Professor in September 2025 and is also an Affiliated Professor at the MIT-IBM Watson AI Lab. OPTML's work spans trustworthy AI, machine unlearning for vision and language models, adversarial robustness, backdoor attack defense in foundation models, and reasoning model safety within the test-time compute paradigm. In May 2025, Open Philanthropy awarded $484,000 to Sijia Liu to support LLM unlearning research aimed at safely removing harmful or misleading information from large language models in ways resistant to jailbreak recovery. The group also receives funding from the Center for AI Safety, Schmidt Sciences, DARPA, Amazon Research, Cisco Research, and national laboratories including LLNL and DSO National Laboratories. Notable recent milestones include a publication on machine unlearning for LLMs in Nature Machine Intelligence (2024), six papers accepted at ICLR 2026, and an OPTML PhD student (Jinghan Jia) being selected for the Anthropic Fellows Program for AI Safety Research in November 2025. The CSE department is hosted within MSU's College of Engineering and is the largest academic unit in that college.
Theory of Change
Updated 05/18/26By developing rigorous techniques for machine unlearning, adversarial robustness, and backdoor defense in large foundation models, the OPTML group aims to make deployed AI systems safer and more controllable. The core bet is that technical safety research — particularly methods to reliably remove dangerous or incorrect knowledge from models and resist adversarial manipulation — reduces the risk that powerful AI systems cause harm, and that publishing this research widely diffuses safety improvements across the AI development ecosystem.
Grants Received
Updated 05/18/26Projects
Updated 05/18/26Research group in MSU’s Computer Science and Engineering department developing optimization‑driven, robust and explainable AI methods to make large‑scale AI systems more trustworthy and scalable.
Discussion
No comments yet. Be the first to share your thoughts.