Purdue University is a major public research university in West Lafayette, Indiana, whose computer science department has received AI safety funding for research on language model robustness and adversarial deception detection.
Purdue University is a major public research university in West Lafayette, Indiana, whose computer science department has received AI safety funding for research on language model robustness and adversarial deception detection.
People
Updated 05/18/26Samuel Conte Professor of Computer Science
Funding Details
Updated 05/18/26- Annual Budget
- $3,159,000,000
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 05/18/26Purdue University is a public land-grant research university founded in 1869 and located in West Lafayette, Indiana. It is the flagship campus of the Purdue University system and is classified as an R1 doctoral university with very high research activity. The university enrolls over 39,000 students and employs more than 10,000 faculty and staff across its West Lafayette campus. Purdue has strong research programs in engineering, computer science, agriculture, and the sciences. In fiscal year 2024-25, the university reported $656 million in research expenditures, a 9% increase over the prior year. Its endowment stood at approximately $4.1 billion in FY2024, and the systemwide budget for FY2025 was approximately $3.2 billion. In the area of AI safety, Purdue's Computer Science department has been active in research on AI security, adversarial robustness, and deception detection. Professor Xiangyu Zhang, a Samuel Conte Professor specializing in AI security, software analysis, and cyber forensics, received a $170,000 grant from Open Philanthropy in December 2022 to support research on improving the robustness of language models against adversarial attacks and detecting deceptive content in NLP applications. This work involves developing model interpretation and gradient-based optimization techniques to identify deceptive outputs generated by humans or AI systems. Purdue also hosts AI Safety Purdue (aispurdue.org), a student-led community of technical and policy researchers aimed at reducing AI risks. This group runs an 8-week introductory AI safety seminar covering alignment, interpretability, and RLHF, as well as a technical reading group. In 2025, Purdue unveiled a comprehensive university-wide AI strategy, including an AI working competency graduation requirement starting fall 2026, making it the first university to mandate such training. Beyond AI safety specifically, Purdue hosts the Institute for Physical Artificial Intelligence (IPAI), described as the nation's first research institute dedicated to advancing AI in the physical world, with focus areas including robotics and supply chain optimization.
Theory of Change
Updated 05/18/26Purdue's AI safety-relevant research, particularly Prof. Zhang's work, focuses on technical solutions to adversarial robustness and deception detection in large language models. The causal chain is: develop interpretability and gradient-based techniques to detect adversarial inputs and deceptive outputs in LLMs, which reduces the risk of AI systems being manipulated or producing harmful deceptive content in high-stakes applications, thereby contributing to safer deployment of AI systems. At the broader institutional level, Purdue's AI safety education programs aim to train a new generation of researchers and practitioners who approach AI development with safety-conscious methods.
Grants Received
Updated 05/18/26Projects
Updated 05/18/26
Community of technical and policy researchers at Purdue University aimed at reducing the risks of advanced AI.
Discussion
No comments yet. Be the first to share your thoughts.