Two recent stories have travelled widely across the internet. In one, a technology founder with bone cancer enters remission after building a personalised treatment strategy, aided in part by artificial intelligence. In another, a man helps develop an experimental cancer vaccine for his dog using AI tools, with apparent success.
At a time when 40 Million people per day rely on AI for health advice, it seems like the answer to any question is just a click away. However, many doctors warn against using ChatGPT for health advice since it can be misleading, unreliable and contain wrong information. For example, a man in his 60's took dangerous advice from AI, swapping salt for another substance called sodium bromide, which caused a rare condition called bromism.
And yet artificial intelligence is placing more scientific information in the hands of those dealing with illness, enabling them to act more independently rather than waiting for medical authorities. This is especially important in countries where health services are stretched and doctors are in short supply.
From passive patient to active participant
Both cases share a common structure. Faced with limited options, the patient or owner refuses to remain passive.
Sid Sijbrandij, the GitLab co-founder and billionaire, did not accept the constraints of standard protocols after his bone cancer returned. Instead, he built a decentralised network of experts, used genomic analysis to understand his disease and explored experimental therapies. AI tools helped him navigate the overwhelming volume of medical research and coordinate insights across disciplines.
He has been in remission from his cancer since 2025.
In the case of the sick dog 'Rosie', the owner, Paul Conyngham, used similar tools to interrogate the scientific literature, identify potential immunotherapy targets and develop a bespoke mRNA vaccine in collaboration with researchers.
In both instances, AI functioned as a cognitive amplifier. It allowed non-specialists to engage more deeply with complex biomedical knowledge than would previously have been possible. It also allowed Paul Conyngham to find the quickest way to produce a vaccine. Usually, an ethics board would stand in between experimental treatments and the application, but because this was an animal, it was easier to do than for a human patient.
This is the real shift: patients are no longer confined to the role of recipients. They are becoming participants and citizen scientists.
The risk of solely relying on AI
Yet those positive stories carry a risk. As oncologists have warned in response to the dog vaccine case, the public takeaway is often not ‘AI can assist experts’, but ‘AI can replace them’. The distinction is not a minor one.
When individuals begin to rely on AI in place of medical guidance, the first problem is an overestimation of what these systems can actually do. AI can summarise research and surface relevant studies, but it does not understand them as a trained clinician does. It cannot properly weigh conflicting evidence, interpret nuance within a clinical context, or take responsibility for the consequences of a decision.
That has not stopped some, with the CEO of the largest hospital group in the US, Mitchell Katz, being quoted as saying, ‘We could replace a great deal of radiologists with AI at this moment, if we are ready to do the regulatory challenge’.
In clinical practice, AI is increasingly being deployed not to replace but rather to support diagnostic work. Researchers at Seoul National University Hospital have developed a tool capable of identifying osteoporosis risk from a routine chest X-ray, highlighting the potential for earlier and more accessible detection.
How AI has changed diagnostics and drug manufacturing
Despite concerns over the risks of using artificial intelligence in diagnosis and preliminary care, adoption of the technology across healthcare is accelerating. An OpenAI report identifies the medical and pharmaceutical sector as one of the fastest-growing areas for AI uptake.
Meanwhile, pharmaceutical company AstraZeneca has partnered with Pangaea Data to develop AI systems designed to detect rare and hard-to-diagnose diseases. The initiative aims to uncover previously undiagnosed patients, improve enrolment in clinical trials and enable more precise treatment strategies.
Artificial intelligence is rapidly reshaping pharmaceutical research, enabling scientists to identify promising drug candidates far more quickly by analysing vast datasets and simulating molecular interactions. This is helping to cut years from development timelines, reduce costs and improve the odds of success in an industry where most experimental drugs fail.
The shift is now global. Large drugmakers are embedding AI across research and clinical development, while forging multibillion-dollar partnerships with specialist firms. China has emerged as a central player in this transformation, with AI-driven biotech companies attracting major Western investment and producing a growing share of early-stage drug discoveries.
One of the clearest signs of this shift is Eli Lilly’s recent deal with Hong Kong-listed Insilico Medicine, which could be worth up to $2.75 billion, among the largest AI-driven drug discovery partnerships to date. The agreement gives Lilly global rights to develop and commercialise medicines designed using Insilico’s AI platforms, underlining how Big Pharma is increasingly relying on technology developed in China.
Optimistic outlook for healthcare and medical research
Whether this model proves sustainable will depend on whether AI-generated candidates can deliver consistent results in late-stage trials, where most drugs still fail. Regulators are also beginning to grapple with how to assess medicines developed with significant algorithmic input, raising questions about transparency, data quality and oversight.
At the same time, AI is beginning to empower patients themselves. In some cases, individuals using publicly available tools and data have identified potential diagnoses or even highlighted overlooked treatment options, feeding back into formal research and accelerating discovery. While this shift carries risks, including a rise in self-diagnosis and misinformation, many in the industry view it as an acceptable trade-off for broader access to knowledge and faster innovation.
For now, however, the direction of travel is clear. As competition intensifies and development costs continue to rise, the ability to move faster from discovery to market is becoming a decisive advantage and one that increasingly hinges on advances in artificial intelligence.
