1. LLMs used for therapy often stigmatise mental health conditions

A study by a team of US researchers concluded that large language models (LLMs) should not replace clinical therapists because their responses often stigmatise mental health conditions and respond inappropriately by, for example, encouraging delusional thinking. It found, for instance, that Meta’s llama3.1-405b model showed stigma 75 per cent of the time.

2. Healthcare sector has high hopes for GenAI, though most projects end at pilot stage

Only 30 per cent of AI projects in healthcare organisations make it past the pilot stage, according to a survey conducted by Amazon Web Services (AWS), Bain & Company and Bessemer Venture Partners. However, an overwhelming majority of organisations – 95 per cent – believe that GenAI will be transformative for the sector, with 84 per cent expecting it to affect clinical decisions and 80 per cent expecting it to reduce labour costs.

3. US healthcare organisations are still mostly reactive on cybersecurity

Healthcare organisations in the US tend to take a reactive approach to cybersecurity, according to the 2025 KLAS Research Healthcare Cybersecurity Benchmarking Study. The study also found that organisations have room for improvement in how they manage third-party risks and assets, with just 53 per cent covering asset management and 52 per cent managing supply chain risks. This year’s study for the first time also looked at management of AI-related risks and found that just 13 organisations responded to note that they are in the early stages of establishing cybersecurity governance in this area.

4. Global survey reveals high hopes for MedTech innovation

A global survey by Johnson & Johnson found that a majority of adults and healthcare providers believe that medical technology innovations will help improve patient care, outcomes and support for older people – for example, 98 per cent of healthcare providers believe that MedTech will help to identify diseases in their earliest stages and 72 per cent of adults believe that innovative technologies will be needed to enhance care for the world’s ageing population.

5. Survey: Health technology challenges include interoperability, security

Although most healthcare executives welcome the benefits of digital health technologies, 91 per cent see interoperability as a challenge and 59 per cent believe that will be a difficult challenge to solve, a survey by the MIT Technology Review sponsored by Roche found. Thirty-eight per cent of respondents agreed that another challenge will be balancing the security of such technologies with usability.

6. 57% of US patients support AI in exam room if it reduces doctors’ time spent on documentation

Fifty-seven per cent of US patients say they support the use of AI to document interactions in the examination room if that means that their doctors can spend more face time with them, according to a survey conducted by healthtech company ModMed. A majority of respondents (77 per cent) note that they spend less than 15 minutes with their doctors in the exam room, and 32 per cent say that doctors spend 7–12 minutes of that time on documentation.

7. Use of AI by healthcare organisations rises to 81% in 2025

The use of AI by healthcare organisations has increased from 61 per cent in 2024 to 81 per cent in 2025, according to a study by the business mobility company SOTI. The survey of 1,750 healthcare decision makers across 11 countries found that the top uses of AI include processing and analysing medical data (60 per cent), updating patient records (59 per cent) and diagnosing medical conditions (40 per cent).

8. Imperial College London: 7 technologies are key to reshaping healthcare in the UK

Seven technologies could help to reshape the NHS and improve healthcare across the UK, according to a white paper from the Institute of Global Health Innovation at Imperial College London. Those technologies include connected data, the NHS app, GLP-1 obesity medications, wearable technologies for remote monitoring, genomics and personalised medicine, AI and robotics for precision care.

9. Biased training data for AI has harmful impacts on different populations

Biased training data for AI models is leading to real-world harms for women, Black people and other populations, say researchers at the University of Birmingham. Addressing such biases will require not just technical solutions but resolving structural inequalities in society and healthcare systems, say Xiaoxuan Liu and Joseph E. Alderman. For example, Liu highlighted one algorithm that underestimated healthcare needs for Black patients because it was based on healthcare spending data. “Contextually, this is not because they have less health needs, it’s because they could afford it less,” Liu says.

10. Many Europeans see tech as a solution to deteriorating healthcare systems

Forty-one per cent of people across France, Germany, Spain and the UK say that they believe their healthcare systems have become worse since the COVID-19 pandemic, according to a survey commissioned by Roche and conducted by GWI. To improve healthcare, 49 per cent say they are willing to share personal data and the same percentage say they are ready to embrace digitalisation for efficiency. Among patients’ top priorities, which could be helped with technological innovations, are early detection and diagnosis of symptoms (51 per cent) and access to new treatments and medical advances (40 per cent).