Professor Elissa Redmiles, Georgetown Security Lab Receive Microsoft Gift for AI Security Research
Elissa Redmiles, the Clare Boothe Luce Assistant Professor in the Department of Computer Science, has received a $25,000 gift from Microsoft for her research on mitigating AI-generated non-consensual intimate imagery.
The gift will support Redmiles’ work with Georgetown Security Lab as part of Microsoft’s External Red Team Alliance, a formalized global extension of the technology company’s AI Red Team. AI red teaming is a testing process that aims to mimic the behavior of someone prompting a model to uncover potential avenues of misuse, Redmiles said.
Georgetown Security Lab is one of 18 university labs across six continents that has received these unrestricted gifts from Microsoft’s AI Red Team.
“Something that’s really nice about this is it’s a fully independent gift, meaning they’re just supporting the work that we are already doing,” said Redmiles, who was named one of MIT Technology Review’s innovators under 35 last year.
Redmiles’ portfolio of research on security for sexual content and interactions includes a focus on addressing AI generation of non-consensual intimate imagery — colloquially known as sexual deepfakes — while protecting free expression. She has collaborated with several Georgetown colleagues, including computer science Professors Sarah Adel Bargal and Yoshi Kohno; postdoctoral fellows Lucy Qin, Eric Zeng and Rupayan Mallick; and Ph.D. student Wisdom Obinna.
Redmiles co-leads a research collective called SafeDigitalIntimacy.org, where a community of computer scientists, psychologists and community experts work together to reduce harm from non-consensual intimate imagery.
“We do research on protecting lawful expression, so people using AI to make sexual content that isn’t abusive and it’s not depicting anyone in particular,” Redmiles said. “And then we also do research on how perpetrators are trying to use AI to generate non-consensual intimate imagery that depicts a real person without their consent and what we can do technically and socially to stop that.”
No solution is perfect, she said, but the multi-pronged projects have focused on the role of AI training data, social norms, deterrence messaging and analyzing the tools that are publicly available. As an example, Redmiles cited research conducted in 2024 on online nudification platforms — or face swap apps.
“We found a large number of face swap apps in the Apple App Store and in the Google App Store could be used for nudification,” she said. “As a result of that work, a lot of those apps got taken down.”
Microsoft’s gift will support Redmiles and her team in continuing this work.
“It’s a nice way to create dialogue between us and industry,” she said.