The summer program has allowed us to get a head start on the upcoming phase of our LeakPro project pertaining to attacks on text-based AI models. The students have explored various possible directions and distilled a report outlining the current state-of-the-art, shortcomings of current approaches, especially in regards to benchmarking datasets, and even presented their own, novel, algorithm that is competitive with state-of-the-art attacks. This report will serve as the starting point and guiding source for the next phase of our project.