Title: Can artificial intelligence models be trusted? Determining the variables for building accurate prediction models for antimicrobial resistance Artificial intelligence (AI) has the potential to be a powerful tool in the field of antimicrobial resistance (AMR), owing to its ability to analyze huge amount of data in no time and predict patterns based on past data. But as they say, with great power comes great responsibility. Before designing and implementing a prediction model in a real-world patient care setting, we need to address two questions. First, what parameters to include before building the model and second, how accurate the predictions are. We aim to address these questions using the global datasets available. Our primary objective is to design multiple AI models to predict susceptibility patterns and determine factors and variables that can be used to validate a prediction model. For validation, we will analyze intra and inter-dataset concordance between predicted result and true result. For intra-dataset analysis, we will build prediction models using a part of the dataset and then apply the model on the remaining part to determine the accuracy. This will help address the following question whether a model built on global dataset can be applied to any healthcare setting or should models be tailored to a specific healthcare setting, geographical areas, population subgroup, wards etc. Our secondary objective is to select the best prediction model and design a web application based on it. We plan to design the prediction results in a minimalistic user interface and make the results more objective, accompanied by a % probability, keeping in mind their application in tertiary care centers, especially in resource-limited settings. We are requesting all the datasets as both intra and inter comparison of datasets are part of our proposal. We believe this will add insights to close the gap between building models and implementing them in real world and thus, contribute to antimicrobial stewardship.