)
)
Welcome to BMJ Future Health.
A patient dies following the use of an AI chatbot. You are their doctor. Are you responsible?
The answer right now is, frustratingly: it depends. A WHO Europe survey published in 2025 found that only four of 50 WHO European member states had liability standards clarifying who is responsible when an AI system causes harm. For patients who follow AI advice and come to harm, without a clinician involved, it is even less clear. Lawsuits are currently before US courts attempting to bring clarity.
"The question of liability hasn't yet been resolved, and there's no clear consensus at present," Henrietta Hughes, England's patient safety commissioner, told The BMJ.
That missing consensus dominated debates this month across the latest BMJ Future Health webinar and a misinformation panel at HLTH Europe 2026.
Also in this newsletter: a new paper outlines why post-launch monitoring should be built into the AI lifecycle from the start. Researchers from Google Research and Google DeepMind selected BMJ Best Practice as one of only two core clinical guideline resources, alongside NICE guidance, to develop and evaluate AMIE, a conversational AI system designed to support clinical management. This study was published in Nature.
We're looking for healthcare professionals who have spotted a problem they think health technology could solve. Could that be you? SUBMIT YOUR PROBLEM BY 4 AUGUST
Misinformation: how our information environment can influence health
Catherine Schuster-Bruce, BMJ Future Health community editor, spoke on a panel on misinformation at HLTH Europe in Amsterdam on 18 June. Misinformation (false or inaccurate information) can influence health outcomes when we trust it and alter our behaviour. In fact, during a keynote at BMJ Future Health Europe 2025, Kedar Mate, co-founder and chief medical officer at Qualified Health, said that by 2030, your information environment (the podcast you listen to, the creator you trust, your doctor, and so on) will shape health outcomes as much as proximity to a hospital or income.
AI chatbots have also entered the information ecosystem. When Schuster-Bruce asked the HLTH Europe audience how many use AI chatbots for health information, around a fifth raised their hands. Randal Whitmore, chief product officer of Ada Health, said (anecdotally) that people increasingly volunteer their personal medical histories to AI systems.
And this is not without risk. Clive Flashman, chief digital officer at Patient Safety Learning and director of strategy at ORCHA, said that patients who followed AI health advice were five times as likely to experience harm as those who did not, citing a Canadian Medical Association survey published in February 2026. In that same survey, 97% of doctors reported that they had personally stepped in to prevent clinical harm after a patient followed misleading online or AI advice, he said.
One audience member asked the panel who carries the blame when misinformation from AI harms a patient.
The answer: Good question! And it remains a fundamental question for clinicians, developers, regulators, and patients alike.
Question of the month
Our next webinar asks: How do you build your career in digital health without leaving your day job?
What has been the biggest barrier: time, credentials, or knowing where to start?
When: 3 July (noon to 1 pm BST)
Join in: Reply and we can put your question to the panel.
Helen Surana, co-editor-in-chief of BMJ Innovations, hosted a webinar on 3 June with Jenny Shand, Paul Wicks, and Gabriella Spinelli to explore the tension between innovation and patient safety.
Key takeaways from the discussion
-
Gabriella Spinelli said that the regulatory journey should be clear, accessible, and affordable. The current process is a particular challenge for small innovators without in-house regulatory expertise, she added.
-
Paul Wicks said that few people can explain how AI tools work. “How they worked in January is not the same as how they worked in June, right?” he said.
-
Jenny Shand said that keeping “clinicians in the loop” is the current status quo for ambient voice technology, but this is not without challenges. The technology can transcribe conversations in different languages, but the clinician must still sign off on the final output. She questioned whether the position changes if the clinician's first language is the language they are speaking with the patient, which might not be English.
Stay informed in a rapidly evolving landscape
News Editorial
- Restoring certainty to global health regulation
- Drift isn't a bug, it's the work: post-deployment surveillance for clinical artificial intelligence
- Meeting clinicians where they are through signal-guided AI innovation
- Analysis of the challenges in ensuring robust and effective clinical evaluation of AI-based digital medical devices
- From bedside to bench: towards clinical predictive AI research that achieves real-world impact
The questions in this issue (about trust, accountability, and what AI looks like in practice) are ones our community is working through in real time. If this prompted a thought, a disagreement, or a question of your own, let us know. We read every response.
Thank you for reading!