Attackers will often target your AI models to extract training data, information about the model and prompt, and even data on on other users.
