A BMJ Open study published in April 2026 found that five widely used AI chatbots frequently produce inaccurate or incomplete medical information. Researchers probed ChatGPT (OpenAI), Gemini (Google), Grok (xAI), Meta AI (Meta) and DeepSeek (High-Flyer) using 250 prompts across five topics — cancer, vaccines, stem cells, nutrition and athletic performance — designed to reflect common public queries and misinformation tropes. Overall, 50% of responses were judged problematic (30% somewhat problematic, 20% highly problematic). Open-ended prompts yielded disproportionately more highly problematic answers than closed questions. Model behaviour varied: Grok produced significantly more highly problematic responses while Gemini produced the fewest. Reference quality was poor (average completeness ~40%), with instances of fabricated citations; responses were typically expressed with unwarranted certainty and few caveats. Only two refusals to answer were recorded, both from Meta AI. The researchers noted limitations: the study stressed models with adversarial prompts, sampled five chatbots tested in February 2025, and commercial systems evolve rapidly, but concluded risks remain if chatbots are deployed for public-facing health guidance without stronger safeguards.

A team of about 50 researchers led by Paul Verweij at Radboud University Medical Center has warned of a "silent surge" in drug-resistant fungi in a Nature Medicine paper published April 2026. The researchers — from 16 organisations worldwide — say resistance is rising in pathogens including Candida auris, azole-resistant Aspergillus fumigatus and Trichophyton indotineae. They argue that resistance often originates in the environment, driven in part by widespread agricultural use of fungicides chemically related to medical azoles, and can spread long distances via airborne spores. People with weakened immunity, such as ICU patients, transplant recipients and cancer patients, face the greatest risk; Candida auris bloodstream infections can carry high mortality. The authors set out a five-step plan calling for increased awareness, strengthened surveillance, improved infection prevention and control, optimized antifungal use, and investments in diagnostics and new drugs. They urge the inclusion of antifungal resistance in the 2026 Global Action Plan on AMR and stress the need for coordinated One Health policies linking agriculture, environment and clinical practice.

A preprint analysis and Reuters report this week found that Eli Lilly’s tirzepatide, while producing larger average weight loss than Novo Nordisk’s semaglutide, was linked with greater losses of lean body mass. The Massachusetts data‑analytics firm nference analysed roughly 1,800 tirzepatide users and about 6,200 semaglutide users, finding tirzepatide associated with an average 1.1% greater lean mass loss at three months and about 2.0% more after 12 months. The preprint categorised a “depletive GLP‑1 metabotype” (>20% total weight loss and >5% lean mass loss) in 10.3% of tirzepatide patients versus 6.7% on semaglutide in the first year. Higher doses, longer treatment and baseline musculoskeletal pain correlated with greater lean mass decline for both drugs. Separately, U.S. prescription data for the week ending April 10 show Eli Lilly’s newly approved oral weight‑loss pill Foundayo had about 1,390 prescriptions versus more than 113,000 for Novo’s Wegovy pill in the same period. Market commentary and investor notes circulated April 16–17 highlighted potential competitive and access implications as the two companies press in the global GLP‑1 market.

Belgian biopharmaceutical group UCB said on April 17 it will acquire U.S.-based Neurona Therapeutics for $650 million upfront and up to $500 million in milestone payments, valuing the deal at as much as $1.15 billion. The acquisition, expected to close by the end of the second quarter of 2026, adds Neurona’s clinical-stage cell therapy NRTX-1001 — a pluripotent stem cell‑derived GABAergic interneuron therapy in Phase I/II trials for drug‑resistant mesial temporal lobe epilepsy — to UCB’s neurology portfolio. UCB said the transaction will not change its 2026 revenue guidance but expects adjusted EBITDA to grow in the high-single-digit to mid‑teens percentage range. Neurona’s backers include Fidelity and Schroders Capital; the startup previously raised significant private funding to advance its regenerative medicine programme. UCB described the move as a strategic expansion into advanced therapies aimed at durable, targeted repair of the nervous system for hard‑to‑treat seizure disorders.

New analyses and recent American Cancer Society reports show colorectal cancer deaths are rising among younger Americans and are concentrated among people with less formal education. ACS researchers, using government death records, examined more than 101,000 colorectal cancer deaths in people aged 25–49 from 1994–2023 and found the death rate in that group rose from about 3 to 4 per 100,000. Increases were almost entirely among those without a four‑year college degree: rates for people with only a high school diploma rose from roughly 4 to 5.2 per 100,000, while rates for those with at least a bachelor’s degree stayed near 2.7 per 100,000. Separate national death‑certificate analyses reported similar education gradients. Colorectal cancer has become the leading cancer killer for Americans under 50; incidence in younger adults has been climbing for years. Experts point to disparities in screening, later diagnosis, obesity, diet, physical inactivity and access to care as likely contributors. Researchers also caution that changes in how education is recorded on death certificates and lack of linked clinical data limit firm conclusions about causes.

About one in four U.S. adults — roughly 66 million people — reported using an artificial intelligence tool or chatbot for physical or mental health information or advice, according to a West Health–Gallup Center survey of more than 5,500 adults conducted Oct.–Dec. 2025 and released in mid-April 2026. Most users turn to AI for quick answers or additional information: 71% cited speed or extra research, 67% said they were curious, and many use AI before (59%) or after (56%) seeing a clinician. A substantial minority cited access and cost barriers: 42% wanted help outside normal hours, 27% did not want to pay for a doctor’s visit and 14% said they could not afford one. While 84% of recent AI health users still saw a provider, 14% reported not seeing a clinician they otherwise would have after receiving AI guidance — an estimated 14 million adults. Trust in AI health information is split roughly evenly: about one-third trust it, one-third neither trust nor distrust it and one-third distrust it; only 4% say they strongly trust AI accuracy, and roughly 11% reported encountering advice they judged unsafe.