One in four GPs are now using artificial intelligence (AI) in their work, despite almost none having formal training or guidance from their employers, new research has found.
The largest year-on-year survey of UK GPs on generative AI has revealed an increase in the number of doctors using tools like ChatGPT in their everyday clinical work.
Some 95% of the doctors who said they use generative AI in their work reported no professional training and 85% of them said their employer has not encouraged its use.
Last year, researchers from the same study found one in five GPs was using the technology.
“In just 12 months, generative AI has gone from taboo to tool in British medicine,” said Dr Charlotte Blease, from Uppsala University in Sweden and Harvard Medical School.
“Doctors are using these systems because they help – not because anyone told them to. The real risk isn’t that GPs are using AI; it’s that they’re doing it without training or oversight.”
Please use Chrome browser for a more accessible video player
Why authors may be right to fear AI
Some 35% of the doctors used AI for writing documentation, 27% for differential diagnoses and 24% for treatment or referrals.
Read more from science and tech:
Female stars ‘may leave sport because of online abuse’
TikTok says teens’ safety not at risk from AI moderation
Teenagers plead not guilty to London transport cyber attack
“This should be a wake-up call,” said Dr Blease.
“AI is already being used in everyday medicine. The challenge now is to ensure it’s deployed safely, ethically, and openly.”
The study’s authors, from Uppsala University, Basel University, the Karolinska Institute, the University of Manchester and Harvard Medical School, surveyed 1,005 GPs around the UK.
Please use Chrome browser for a more accessible video player
Could the AI bubble burst?
The findings were published on Tuesday in the Digital Health science journal.
Researchers highlighted the risks of using AI in clinical settings like the technology’s tendency to get things wrong, or “hallucinate”, and the potential for “algorithmic discrimination” because of potential biases in the models’ training data.
The authors also raised concerns about the data privacy of patients.
Sky News has contacted the UK’s NHS bodies for comment.
