This Project Grant award from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program (CFDA 47.070) is providing $400,000 to Northeastern University to conduct research on understanding, visualizing, and attributing multimodal generative AI models. The project aims to develop new methodologies for analyzing the internal mechanisms of these large-scale AI models that can generate both text and images. The research will focus on three main thrusts: 1) visualizing the models' internal structures and their roles in the generation process, 2) exploring algorithms to manipulate the discovered concepts and relationships to address inconsistencies, biases, and safety issues, and 3) developing attribution methods to assess the influence of training data on model outputs. The award, made on Sep 15, 2024, supports this 4-year research effort to increase the accountable and safe use of state-of-the-art multimodal AI models and mitigate potential harms. The project will publicly release all materials, including code, models, and datasets, and engage the research community through tutorials, workshops, and courses.