About Me
I am an Assistant Professor of Computer Science at Dartmouth College, with a Ph.D. from the University of Chicago. My work addresses the most pressing challenges and concerns in AI.
My research examines the technical mechanisms of generative AI systems, with a focus on training-data attribution, model memorization and regurgitation, provenance identification, and auditing methods for generative outputs. My work has been featured in outlets including The New York Times, The Wall Street Journal, Bloomberg, CNN, and TIME.
My contributions have been recognized with the Forbes 30 Under 30 award, MIT Technology Review Innovator of the Year, and MIT 35 Under 35. This research has also earned the USENIX Internet Defense Prize, Distinguished Paper Awards at top conferences, and the Concept Art Association’s Community Impact Award.
I am recruiting 1–2 fully funded PhD students to start in Fall 2027. If you are interested in security and trustworthy AI, apply in the Fall 2026 cycle through the Dartmouth CS system and mention me in your research statement. Please email me your CV beforehand.
News
- We are organizing the first ArtSec Workshop at Oakland'26. Please consider to submit your work.
- Glaze has been downloaded over 7.6 million times in 150+ countries; Nightshade has reached over 2.5 million downloads.
- Named the MIT Technology Review 2024 Innovator of the Year and 35 Under 35.
- Selected for the 2024 Forbes 30 Under 30 list.
- Featured in MIT Technology Review for the Nightshade project.
- Glaze recognized by TIME Magazine as a 2023 Best Invention (Special Mention).
- Awarded the Internet Defense Prize and a Distinguished Paper Award at USENIX Security.
Selected Work
- Identifying Provenance of Generative Text-to-Image Models (USENIX Security 2026): determining which model produced a given AI image.
- Organic or Diffused: Can We Distinguish Human Art from AI-generated Images? (CCS, Distinguished Paper): telling human-made and AI-generated images apart.
- Poison Forensics: Traceback of Data Poisoning Attacks: tracing a model's behavior back to the specific training data responsible.
- Glaze & Nightshade: widely adopted tools that study how changes to training images affect what generative models learn and produce.