arXiv· arXiv:2602.07152· Kristopher W. Reese, Taylor Kulp-McDowall, Michael Majurski, Tim Blattner, Derek Juba, Peter Bajcsy, Antonio Cardone, Philippe Dessauw, Alden Dima, Anthony J. Kearsley, Melinda Kleczynski, Joel Vasanth, Walid Keyrouz, Chace Ashcraft, Neil Fendley, Ted Staley, Trevor Stout, Josh Carney, Greg Canal, Will Redman, Aurora Schmidt, Cameron Hickert, William Paul, Jared Markowitz, Nathan Drenkow, David Shriver, Marissa Connor, Keltin Grimes, Marco Christiani, Hayden Moore, Jordan Widjaja, Kasimir Gabert, Uma Balakrishnan, Satyanadh Gundimada, John Jacobellis, Sandya Lakkur, Vitus Leung, Jon Roose, Casey Battaglino, Farinaz Koushanfar, Greg Fields, Xihe Gu, Yaman Jandali, Xinqiao Zhang, Tara Javidi, Akash Vartak, Tim Oates, Ben Erichson, Michael Mahoney, Rauf Izmailov, Xiangyu Zhang, Guangyu Shen, Siyuan Cheng, Shiqing Ma, XiaoFeng Wang, Haixu Tang, Di Tang, Xiaoyi Chen, Zihao Wang, Rui Zhu, Susmit Jha, Xiao Lin, Manoj Acharya, Weichao Zhou, Feisi Fu, Panagiota Kiourti, Chenyu Wang, Zijian Guo, H M Sabbir Ahmad, Wenchao Li, Chao Chen·Indexed by this site · Originally published
Final report of the Trojans in Artificial Intelligence (TrojAI) program
Trojans in Artificial Intelligence (TrojAI) Final Report
AI summary
The final report of IARPA's TrojAI program summarizes the program's multi-year research on AI trojans, covering detection methods, mitigation measures, and test and evaluation results. The report presents methods for detecting hidden backdoors in models through weight analysis and trigger inversion, as well as approaches to mitigating trojan risks in deployed models. Comprehensive testing and evaluation provide results on detector performance, sensitivity, and the prevalence of "natural" trojans. The report concludes by summarizing program lessons and offering recommendations for advancing AI safety research.