Daily briefing

Hospitals race to patch a Java flaw found with AI in an hour

Nov 20, 2026

Illustration: a cracked coffee mug with steam rising from it has a bandage placed over the crack.

Hospitals are racing to patch a critical Java flaw that researchers said AI helped them find in about an hour, after a federal cybersecurity agency urged immediate fixes Nov. 17.

Oracle released a fix the same day, and hospital cyber teams are scanning networks and medical devices to "patch before they hack," the American Hospital Association's cybersecurity adviser told Becker's. A JAMA cohort of 3.7 million children tied each head CT to about one extra brain tumor per 1,266 scans.

FDA warned that a B. Braun needle-free IV valve can stay open after disconnection, and the U.S. Preventive Services Task Force met for the first time in more than a year.

  • ActIncident Nov 17, 2026

    CISA urges organizations to find all Java installations and apply Oracle's Nov. 17 patches immediately, internet-facing systems first.

    HospitalsPractices
    The Cybersecurity and Infrastructure Security Agency urged organizations on Nov. 17 to patch Java4Shell (CVE-2026-114938), a memory-corruption flaw in the Base64 decoder of Java's x86-64 HotSpot virtual machine that can let an attacker run code remotely without logging in. Oracle fixed it in its November security update the same day. Researchers at the security startup Calif said AI let them find the bug in about an hour and build a working exploit in two more for less than $200 in AI tokens, The Record and Becker's reported. CISA told organizations to look for Java bundled inside third-party applications, container images and appliances and to monitor for unexpected JVM crashes and rogue runtimes. Hospital cyber teams are scanning internal networks, medical devices and internet-facing programs to "patch before they hack," said John Riggi, the American Hospital Association's national cybersecurity adviser; the association said it knew of no effect on hospital operations beyond the rush to deploy defenses, and Epic said its systems were not affected, according to Becker's.
  • FYIEvidence Nov 19, 2026
    Brain tumor risk rose with cumulative radiation dose from medical imaging in a retrospective cohort of 3,704,860 children born from 1996 through 2016 in six U.S. health systems and Ontario, JAMA published Nov. 19. Over 45.7 million person-years, 1,536 brain tumors were diagnosed, and one head CT delivered a mean 66.2 mGy to the brain. The excess relative risk was 1.27 per 100 mGy (95% CI, 0.83-1.77), and hazard ratios against no exposure rose from 1.86 at 5 to less than 50 mGy to 3.85 at 150 mGy or more. Exposure comparable to one head CT was associated with 7.9 excess brain tumors per 10,000 children by age 21, about 1 per 1,266 head CT examinations, and imaging accounted for an estimated 10.1% of the tumors. The authors said the findings support careful clinical justification and dose minimization when pediatric imaging with ionizing radiation is indicated.
    Sources: JAMA, PubMed
  • ActIncident Nov 19, 2026

    FDA and B. Braun say to inspect Caresite Micro valves after each disconnection and replace any that leak or stay open.

    FDA posted an early alert Nov. 19 that B. Braun Medical's Caresite Micro Luer Access Device, a needle-free valve used for blood draws, IV therapy and medications, can keep its internal piston depressed after a male luer is disconnected, leaving the valve open. The open valve can cause backflow and leakage and, less often, air ingress, with risks of blood loss, interrupted therapy, medication loss, air embolism and microbial contamination; neonates, children, critically ill patients and those on life-sustaining therapy face the highest risk, FDA said. The company linked the problem to missing silicone lubricant at the piston. B. Braun told customers in a Nov. 13 letter to inspect each device after disconnection and replace any that leak or stay open, replacing the whole extension set when the valve is bonded to one. As of Nov. 3, the company had reported no serious injuries or deaths, FDA said.
    Source: FDA
  • WatchRegulation Nov 19, 2026

    The Nov. 19 agenda listed ratification of final cervical cancer, alcohol use and perinatal depression recommendations; no publication dates were given.

    Primary care
    The U.S. Preventive Services Task Force, which recommends which preventive services must be covered by insurance, convened Nov. 19 for its first meeting in more than a year, after Health Secretary Robert F. Kennedy Jr. canceled its meetings last year and overhauled its membership, STAT reported. The first day's agenda listed ratification of final recommendations on cervical cancer screening, screening and counseling for unhealthy alcohol use in adolescents and adults, and counseling to prevent perinatal depression, topics that have not been updated in at least five years and have awaited finalization since last year, according to STAT. The panel also heard an evidence presentation on diet, physical activity and weight loss to prevent cardiovascular disease. Its Nov. 20 session takes up chronic kidney disease, colorectal cancer and lung cancer screening.
  • FYIEvidence Nov 19, 2026
    All physicians
    Adults rated physicians who disclosed using AI for administrative work more favorably, and were more willing to seek their care, than physicians who disclosed using it for diagnosis or treatment, npj Digital Medicine published Nov. 19. The 5-by-2 factorial experiment randomly assigned 1,574 adults in mainland China to physician profiles that varied the kind of AI use, including an explicit statement that no AI was used, and the severity of the patient's condition. Severe conditions amplified negative responses to therapeutic AI, and the explicit no-AI disclosure was associated with greater willingness to book an appointment when the condition was severe. A community sample of 128 adults in Macao showed directionally consistent main effects, and the authors called for preregistered replication using behavioral outcomes.
  • FYIEvidence Nov 17, 2026
    Patients trusted advice from an AI skin cancer app more, and were more likely to say they would follow it, when a teledermatologist supervised the result, JMIR mHealth and uHealth published Nov. 17. In the online vignette experiment, 879 Dutch adults from the Erasmus MC patient panel (mean age 62.5) judged simulated app outputs; trust scored 76.9 of 100 for supervised advice versus 65.7 for AI-only advice, and the odds of intending to follow the advice were 1.7 times higher (95% CI, 1.28-2.22). A high-risk result also raised adherence intentions (OR, 2.9), and the effect of supervision on adherence was fully mediated by trust, the authors reported. Participants rated simulated outputs, not their own care.

Latest news

  1. Medicare Advantage plans would have to make in-network clinicians file prior authorizations through an electronic interface, CMS proposes
  2. Medicare's first mandatory Advantage model would swap star ratings for local 'Diamond Ratings' in randomly chosen states from 2029
  3. A California health system says an employee uploaded emergency patients' diagnoses and record numbers to ChatGPT

See all the latest news