Tag
This study demonstrates that large language models with institution-specific prompting can recover protected health information missed by existing de-identification systems, enhancing data privacy compliance in electronic health records.