Meet us at SABCS 2026 December 8-11, 2026 - San Antonio

Why a slowdown on AI could cost patients their lives

“What happens to patients while we wait? ”. In an opinion piece for Medical Economics, published November 18, 2026, xCures CEO Mika Newton makes the case against a blanket slowdown on AI in medicine.

Key points

  • Today’s harm is already measurable. Johns Hopkins researchers estimated in 2023 that diagnostic error kills or permanently disables 795,000 Americans every year. A slowdown keeps that baseline in place.
  • The evidence already exists. An AI sepsis early warning system tracked 590,736 patients across five hospitals. When a clinician confirmed the alert within three hours, sepsis mortality was 18.7% lower (Nature Medicine).
  • Screening shows the same pattern. In a Swedish randomized trial of 105,934 women, AI-supported mammography reading found 29% more cancers with no rise in false alarms. Screen-reading workload dropped 44% (The Lancet Digital Health).
  • Not all AI is the same. The FDA has authorized more than 1,500 AI-enabled medical devices. Most handle narrow, supervised tasks like spotting strokes on CT scans. A blanket slowdown can’t tell these apart from frontier AI models.
  • Skeptics inside medicine have a fair point. Faulty alerts and noisy tools are real, and hospitals are right to scale back or rebuild them. Pew found 72% of US adults say it’s extremely or very important that providers tell them when AI is used in their care.
  • He proposes a clear test. Judge each tool on its own results in real clinical settings, then keep the ones that help and drop the ones that don’t. Trust, he argues, rests on published evidence, open disclosure, and traceable sources.


xCures® CEO Mika Newton argues that medical AI should be judged one tool at a time, on published evidence, rather than paused as a single category.

Read the full article in Medical Economics →