Exploring the limits of strong membership inference attacks on large language models

Open in new window