Membership inference for language models investigates the ability to determine if specific data points were part of a model's training set. This research focuses on identifying vulnerabilities in models that allow adversaries to infer training data membership, developing techniques to measure and mitigate these risks.