A recent study in the UK shows that nearly 30% of general practitioners (GPs) have used artificial intelligence tools including ChatGPT during face-to-face consultations with patients to generate consultation summaries, assist with diagnosis and handle daily administrative affairs. However, at the same time, they are also generally worried about the risk of misdiagnosis and legal liability issues caused by this.
This study, conducted by the think tank Nuffield Trust, is based on a survey of 2,108 family doctors on the application of AI by the Royal College of General Practitioners, and combined with multiple focus group interviews, it outlines the rapid expansion of AI in the UK's primary care system but the lack of supervision. The study pointed out that high-level government officials are counting on AI to alleviate patients' difficulties in making appointments and shorten waiting times. However, AI applications in primary medical practice are more like growing spontaneously in a "regulatory wilderness", and it is often difficult for doctors to judge which tools have been formally evaluated and approved at the national or NHS level.
The survey shows that among the 2,108 doctors who participated in the survey, 598 (accounting for 28%) have stated that they have used some form of AI tools in their work. The usage rate of male doctors is 33%, which is significantly higher than the 25% of female doctors. The proportion of clinics using AI in wealthy areas is also much higher than that in poor areas, reflecting the gender and geographical imbalance behind the so-called "technology diffusion". At the same time, regardless of whether AI has been used, the vast majority of doctors interviewed expressed similar concerns: Once relying on AI to assist decision-making, medical institutions and individual doctors may face higher professional liability and medical and legal risks when errors occur, and patient privacy and data security are also regarded as key risks.
Becks Fisher, director of research and policy at the Nuffield Trust and a general practitioner, pointed out that the government has high hopes for AI at the macro level, but the reality is that there is a lack of unified standards and systematic training in grassroots implementation, resulting in many doctors "seeing the potential but not knowing how to use it safely." Currently, NHS integrated care committees in different regions have different policies on the use of AI. Some encourage doctors to adopt new tools, while others explicitly prohibit them, further exacerbating doctors' confusion about compliance boundaries. What is even more disappointing to policymakers is that the survey found that the time saved by AI was not mainly converted into more additional outpatient visits, but was used by doctors for self-adjustment and rest, including reducing overtime to prevent professional burnout, which deviated from the expectations of policymakers to "use AI in exchange for more medical consultations."
Another independent study published in the journal "Digital Health" also confirmed a similar trend: in the past year, the proportion of family doctors using AI increased from 20% to 25%, showing that generative AI is rapidly completing the role transformation from a "taboo topic" to a "routine tool" in British medical practice. The first author of the study, Charlotte Blease of Uppsala University in Sweden, emphasized that the real risk is not that doctors start to use AI itself, but that they "fumble around" in the absence of systematic training and formal supervision. She called on the fact that AI has irreversibly entered daily clinical practice. The next step should be to ensure that its use process is open, transparent, safe, controllable and ethical, rather than dwelling on the question of "whether to use it or not."
At the same time, the use of AI on the patient side is also rising. The latest survey by Healthwatch England, a British patient watchdog, shows that although the public still mainly trusts the source of health information provided by the NHS, about 10% (9%) of the respondents will use AI tools to obtain health care advice, especially when they cannot make an appointment with a general practitioner in time. They are more likely to "ask AI first". However, the quality of advice given by AI varies. In one case, a patient received a reply after using AI consultation that confused herpes zoster with Lyme disease, highlighting the real dangers of patients relying on AI self-diagnosis in the absence of professional review.
Faced with the rapid penetration of AI into the frontline of medical care, the British government has launched a special committee in September to study how to promote AI medical applications while ensuring safety, effectiveness and standardization, and is expected to propose a package of regulatory and practical recommendations in the future. However, before a formal framework is introduced, grassroots doctors still have to carefully balance their expectations of reducing burdens and increasing efficiency with potential errors, privacy leaks, and litigation risks. AI has also gradually changed from a "taboo topic" to a "new tool" that can be used in this swing.