University researchers in Oxford have published a peer-reviewed–style evaluation that tests whether automated tools — both specialist de‑identification software and large language models (LLMs) — can reliably remove patient identifiers from real, routine electronic health records (EHRs), and...