As AI continues to reshape healthcare, staying informed about emerging risks, governance, and practical implementation challenges is becoming an essential part of professional practice. This month, we’ve selected two thought-provoking articles that explore different sides of responsible AI use in healthcare, from organisational governance frameworks to the growing issue of deception in clinical large language models.
Governance for safe and responsible AI in healthcare organisations: a scoping review of frameworks
If you’re trying to make sense of the growing pile of AI governance frameworks out there, this review does the legwork for you — surveying 77 frameworks worldwide and pulling out what actually makes one useful in practice. It offers a genuinely helpful distinction between ethics principles (like fairness and transparency) and governance principles (the practical processes organisations need, like data management and ongoing monitoring) — a distinction many frameworks blur. It also flags some notable blind spots worth knowing about, including how few frameworks address oversight committees, cultural safety, sustainability, or the risks of commercialising AI tools. A worthwhile read for anyone in healthcare trying to build (or implement) a governance approach that will hold up in the real world. Read more.
Deception in clinical large language models: an under-recognised safety risk
You have heard about bias and hallucination, but what about deception? This commentary by Reddy et al in The Lancet Digital Health is a brief primer on deception in clinical large language models — a distinct risk where an AI’s output misrepresents what it can do or why it reached a conclusion, rather than simply getting facts wrong. The authors break it down into three patterns worth recognising: models falsely claiming abilities they don’t have (like checking real-time drug interactions), models fabricating a plausible-sounding rationale to back up a wrong answer, and models simply telling clinicians what they want to hear rather than what’s accurate. With real-world examples of deception, this commentary is a starting point for understanding what deception is and how it is occurring in clinical LLMs. Read more.
Enjoyed a book, blog, video or podcast recently that you think other CHIAs might also like? Tell us what and why by emailing [email protected].