Red-teaming is often cited as a solution for many of the potential problems posed by AI adoption. But it’s not that simple. While it can be a valuable tool in a broader test and evaluation toolkit, red-teaming has many limitations and is far from a silver bullet. 

On March 25, CSET experts Jessica Ji and Colin Shea-Blymer will host Microsoft’s Tori Westerhoff, MITRE’s Christina Liaghati, and Marius Hobbhahn of Apollo Research for a discussion about how AI red-teaming is done in practice and how to make it more useful in the future. They’ll explain how red-teaming fits in with a broader set of test and evaluation tools, and what different stakeholders can do to maximize its benefits.

Participants

Anna Raney is a Lead AI Security Engineer at The MITRE Corporation. She holds a B.S. in Electrical Engineering and M.S. in Computer Engineering, specializing in image processing and pattern recognition, from the University of New Hampshire. Since joining MITRE in 2020, Anna has explored various facets of AI, but found her passion to be AI Security when she led a project on embedding watermarks into object detection datasets for data provenance. She is now the AI Red Teaming & Security Group Lead within MITRE’s AI & Autonomy Innovation Center. In her role, Anna works closely with diverse stakeholders from the Department of Defense (DoD), Intelligence Community (IC), and Department of Homeland Security (DHS), developing AI capabilities that are safe, secure, and trustworthy. Additionally, she contributes to MITRE ATLAS, a comprehensive, globally accessible knowledge base that details adversary tactics and techniques against AI-enabled systems.

Tori Westerhoff is currently a Principal Director and AI Red Teaming Manager at Microsoft, where she leads Microsoft’s AI Red Team’s operators who whitehat-hack high-risk GenAI products pre-launch. With a strong background in both technology and analytics, Tori previously led strategy for Microsoft AR/VR and Metaverse businesses and worked as a People Analytics lead for Microsoft C-suite members, where she created NLP/AI tools to improve workplace metrics. Before joining Microsoft, she worked in national security strategy with Deloitte Consulting, supporting various intelligence, law enforcement, and defense agencies. Tori holds a B.A. in Cognitive Neuroscience from Yale University and an MBA with dual majors in Marketing and Operations from Wharton and served as one of the first members of the Wharton Neuroscience Initiative.

Marius Hobbhahn is the Founder and CEO of Apollo Research, a London-based AI safety organization focused on reducing dangerous capabilities in advanced AI systems, particularly deceptive behaviors. Apollo Research designs model evaluations and conducts interpretability research to better understand state-of-the-art AI models, while providing technical guidance to policymakers. Marius previously worked as a Research Fellow at Epoch, modeling and forecasting AI progress. With a background in machine learning from the University of Tübingen, his expertise spans frontier AI evaluations, Bayesian machine learning, and AI governance. Marius was a participant at the 2023 UK AI Safety Summit at Bletchley Park and has spoken at international AI workshops and panels, including the Beijing Academy of AI.

Colin Shea-Blymyer is a Research Fellow at CSET, where he works on the CyberAI Project. His research has spanned safe reinforcement learning, formal methods, adversarial machine learning, and AI ethics. Previously, he was a graduate researcher with MITRE, where he helped establish the National Institute of Standards and Technology (NIST) program on adversarial machine learning research at the National Cybersecurity Center of Excellence (NCCOE). He holds an MS and BS in Computer Science from Virginia Tech. He has a PhD in Computer Science and Artificial Intelligence from Oregon State University, where he was an Outstanding Scholar Fellow.

Jessica Ji (Moderator) is a Research Analyst at CSET, where she works on the CyberAI Project. Before joining CSET, she worked as a software engineer at Expedia Group. Jessica received an M.A. in Security Studies from Georgetown University and a B.A. in Computer Science from Princeton University.