UK needs new laws for AI in healthcare, says watchdog

Britain’s primary industry watchdog for medical products has issued a stark warning: the United Kingdom urgently requires new, comprehensive regulations to govern the burgeoning use of Artificial Intelligence (AI) within the National Health Service (NHS) and other healthcare settings. The Medicines and Healthcare Products Regulatory Agency (MHRA), the body responsible for scrutinising and approving all medical devices and licensing therapeutic drugs in the UK, has unveiled a pivotal report containing 44 detailed recommendations aimed at modernising its existing policies. This proactive stance comes as the integration of AI technologies into the healthcare sector accelerates at an unprecedented pace.

Lawrence Tallon, Chief Executive of the MHRA, articulated a clear vision for the future, telling the BBC that AI is poised to become a routine element of NHS care delivery. "What I would expect is that patients will increasingly see AI as part of the way that normal NHS healthcare is delivered," he stated. "That should happen in a way that they can maintain their trust and their confidence in what’s happening." This emphasis on patient trust underscores the delicate balance between technological advancement and the fundamental principles of healthcare.

The comprehensive report was not a solitary endeavour; it was meticulously compiled by an independent commission that actively sought and incorporated input from a vast swathe of stakeholders. This included the perspectives of over 12,000 individuals, encompassing both patients who directly experience healthcare services and the clinicians who provide them. This broad engagement ensures that the recommendations are grounded in real-world experiences and address a wide spectrum of concerns.

The MHRA’s recommendations are designed to bridge the gap between outdated regulatory frameworks and the dynamic nature of AI. The existing medical devices regulatory framework, Tallon explained, was predominantly conceived during an era focused on regulating physical implants like hip and knee replacements, or simpler instruments such as stethoscopes and plasters. While these guidelines may prove adequate for rudimentary AI applications, such as those trained to identify known symptoms on medical scans, they fall short when it comes to more sophisticated and adaptive AI models.

A critical distinction highlighted by Tallon is the inherent characteristic of AI: its continuous evolution post-authorization. "Unlike most of the medical products we’re used to regulating, these products continue to change after the point of authorization," he explained. "As new data gets fed in, they learn, they adapt, they drift." This dynamic nature presents a significant challenge for traditional regulatory approaches, which typically assess products based on a static design. The MHRA’s proposed updates aim to establish mechanisms for ongoing monitoring and re-evaluation of AI systems as they evolve.

Acknowledging the global nature of this challenge, Tallon admitted, "I don’t think at this moment in time we can point to a single country, a single regulatory framework, and say that they have absolutely cracked it." This admission highlights the need for international collaboration and shared learning as countries grapple with the complexities of AI regulation in healthcare.

The immediate impact of AI in healthcare is already being felt. AI-powered "scribes," also known as note-takers, which leverage the capabilities of Large Language Models (LLMs), are reportedly being utilised by a significant 40% of UK-based General Practitioners (GPs) to efficiently record patient consultations and generate comprehensive reports. This technology promises to alleviate administrative burdens on clinicians, allowing them to focus more on patient care.

However, this widespread adoption is not without its concerns. A recent study conducted by the University of Edinburgh shed light on a potential drawback: patients may exhibit a reduced willingness to disclose sensitive personal information, such as their history of substance abuse, if they are aware that their conversations are being processed by AI. This finding underscores the importance of transparency and patient consent in the deployment of AI.

Professor Henrietta Hughes, a GP who was an integral part of the report commission, shared her observations from patient interactions. She noted that while many patients express contentment with the use of AI during their consultations, a segment chooses to opt out. "Some say, ‘I don’t want to talk to a robot,’ and that is also fine," she remarked, emphasizing the necessity of respecting patient autonomy and offering alternatives.

Professor Hughes also addressed the potential for errors in AI-generated notes, acknowledging that scribes can indeed make mistakes. However, she firmly placed the responsibility for accuracy on the shoulders of the doctors. "It was the responsibility of doctors to correct them," she asserted, reinforcing the concept of human oversight and accountability in AI-assisted healthcare.

While current AI applications in healthcare are predominantly focused on administrative tasks and symptom identification under human supervision, the excitement surrounding its potential in health and medical research industries is palpable. Professor Alastair Denniston, an ophthalmologist and fellow contributor to the report, lauded AI as an "exceptional opportunity" for healthcare, one that he believes is "likely to rank alongside step-changes such as antibiotics and MRI." This comparison underscores the transformative potential of AI to revolutionise medical diagnosis, treatment, and discovery.

The optimistic outlook extends to ambitious predictions. In September, Rene Haas, the chief executive of chip designer Arm, expressed his conviction to the BBC that AI would be instrumental in discovering a cure for cancer "within our lifetime." Such bold forecasts highlight the immense promise and the high expectations placed upon AI’s capabilities.

Despite the widespread enthusiasm, significant concerns persist. A primary area of apprehension revolves around the potential for AI models to generate incorrect medical advice or make flawed diagnostic decisions. These errors can arise from inherent biases present in the vast datasets used to train these AI systems, leading to skewed outcomes that may disproportionately affect certain patient demographics. The risk of AI chatbots providing inaccurate medical guidance also remains a pressing issue that requires robust safeguards and continuous evaluation. The MHRA’s call for new regulations is therefore a crucial step towards ensuring that the integration of AI into the UK’s healthcare system is both innovative and safe, fostering trust and promoting equitable access to high-quality care. The recommendations aim to establish a dynamic and adaptable regulatory framework that can keep pace with the rapid evolution of AI, thereby safeguarding patient well-being and unlocking the full potential of this groundbreaking technology for the benefit of all.