Healthcare AI News 4/23/25

News

The UK’s health secretary says that AI-driven health checks – called MOTs for measurements, observations, and tests — could transform care for frail patients who are over 65 by using machine learning and genomics to speed diagnosis, guide treatment, and predict illness. Japan offers a similar early detection program called Ningen Dock, a cash-only program that uses imaging, endoscopy, and lab work to generate personalized risk assessments. That service is also offered to foreign residents in a medical tourism package that is covered by some US insurers, such as Aetna (above).

Google DeepMind CEO Sir Demis Hassabis – who won a Novel Prize in chemistry last year after starting his career as a designer of widely popular video games — predicts in a “60 Minutes” interview that AI will reduce drug development time from years to weeks, making all diseases curable within 10 years.

Agentic coding platform Cursor draws online scorn and customer cancellations after its AI support agent incorrectly blames a user’s inability to run multiple sessions as company policy rather than a software bug. Cursor says that it will start labeling AI-generated responses after users questioned whether it was trying to pass off its “Sam” assistant as human. It also fixed the bug that the user had reported.

Business

Just 30% of healthcare AI pilot projects reach production, a new study finds, most often stalled by security concerns. Providers are much more interested in trying and buying AI solutions than they were with EMRs.

Middle East specialty provider Burjeel Holdings will use Hippocratic AI’s agents for patient-facing, non-diagnostic clinical tasks that will be delivered in multiple languages.

Research

A new study finds that AI models beat PhD-level virologists in solving complex wet lab problems, scoring 44% versus the experts’ 22%. While the findings offer hope for advancing infectious disease control, they also raise alarms about AI’s potential use to develop bioweapons.

Stanford Health Care researchers find that endocrinologists view AI-generated draft responses to patient portal messages as helpful, but see tools that use patient data, such as for triage, as risky. AI was rated most useful for administrative tasks like writing authorization letters and patient education, with the greatest potential use being the management of patient scheduling.

Other

A Spain-based Microsoft engineer who was frustrated by repeated misdiagnoses of his son’s rare condition develops DxGPT, an AI tool that analyzes user-reported symptoms to suggest possible diagnoses. Access is free.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Follow on X, Bluesky, and LinkedIn.
Contact us.