Large language models (LLMs) have demonstrated remarkable capabilities in natural language understanding and generation, but recent concerns have emerged regarding their potential to memorize and reveal sensitive information present in their training datasets [7, 12, 14]. This paper delves into the phenomenon of “extractable memorization” in language models, a concept distinct from discoverable memorization, as … Leggi tutto Unveiling Extractable Memorization in Large Language Models: A Case Study on ChatGPT