Here's an uncomfortable question for the future of medicine: What happens when the Wi-Fi goes down and your doctor only knows how to practice medicine with AI assistance?
A group of medical professors from the University of Missouri's School of Medicine is raising this exact concern, warning that artificial intelligence could seriously damage medical students' critical thinking abilities if educational institutions don't establish proper guardrails. Writing in a December 1st op-ed published in BMJ Evidence-Based Medicine, the authors argue that medical education desperately needs an overhaul to address AI's disruptive influence.
"The goal of using AI to augment education, rather than letting it erode independent reasoning, is a worthy pursuit," the professors wrote. "As AI is disrupting traditional learning and evaluation methods, adjustments to medical school and training curricula are necessary."
The problem, according to these experts, is that medical schools currently have "largely insufficient institutional policies and guidance" governing how students use AI in their homework and clinical training. Without proper oversight, future doctors risk becoming so dependent on technology that they lose essential skills. And that's when things get dangerous.
"What happens if the servers or AI services go down?" the op-ed asks pointedly. "The impact of this is particularly ominous for learners who are working on developing the skill in the first place, as they are denied the opportunity to do so in the process."
It's not that AI doesn't belong in medical education—the authors acknowledge its potential value. But students need to learn how to effectively use these tools while maintaining their ability to think independently. That means teaching them to verify AI outputs and recognize when the technology gets things wrong.
"Medical training should include practice in rejecting poor AI advice and in explaining why it is unsafe to follow," the essay states.
The Reality of AI in Healthcare
This isn't a theoretical debate about some distant future. AI has already taken the medical world by storm. According to the American Medical Association, two-thirds of physicians used AI in their practices in 2024, a dramatic jump from just 38% the previous year.
Yet despite this rapid adoption, healthcare actually lags behind other industries when it comes to AI integration. The World Economic Forum noted in a recent report that one major roadblock is "increased distrust" in AI's abilities and effectiveness. And honestly? That skepticism might be justified.
The Missouri professors point to a persistent and troubling issue: AI systems regularly generate false information with complete confidence. "Hallucinating confident falsehoods and sources remains a frequent failure mode for AI models," they wrote.
Research backs up this concern. A study published in Communications Medicine earlier this year found that large language models are "highly susceptible" to producing false and potentially dangerous information in clinical settings. The stakes couldn't be higher—we're talking about technology that could literally make life-or-death recommendations based on fabricated data.
This problem hit close to home earlier this year when a report released by U.S. Health Secretary Robert F. Kennedy Jr. cited studies that simply didn't exist—a stark reminder of how AI hallucinations can infiltrate even high-level decision-making.
Building Better AI-Integrated Medical Education
So what's the solution? The op-ed authors propose a comprehensive approach that balances AI literacy with fundamental skill development.
First, they argue that educators need to assess not just what students produce, but how they use AI to get there. "This can be done by asking the students to 'show their work', provide a paper trail and even submit the LLM prompts they used along with written rationales for accepting or rejecting the AI's output," they explained.
Second, students need regular evaluation in completely AI-free environments to ensure they're developing core competencies. The professors specifically highlight bedside communication, physical examination, teamwork, and professional judgment as areas where technology-free assessment is "especially important."
Finally, medical curricula should include robust AI literacy training. Students don't necessarily need to understand the deep technical minutiae of how these systems work, but they should grasp the fundamental principles behind AI's capabilities and limitations.
"Medical trainees may not need to be fully emerged into the technical data engineering details and training pipelines for AI models," the authors wrote, "but they should understand that process in principle and grasp the concepts underpinning its strengths and weaknesses."
The bottom line? AI isn't going anywhere in healthcare, but the next generation of doctors needs to be trained to use it wisely—not become dependent on it blindly. The goal is augmentation, not replacement of human judgment.
