Why Doctors Are Pushing Back Against Claims That AI Can Replace Them
Chicago, Saturday, 5 September 2026.
Despite 66% of US doctors using clinical AI, the American Medical Association strongly rejects a new study claiming standalone algorithms will soon deliver superior patient care without human oversight.
Study Claims AI Superiority Over Human Physicians
A new paper published in the Journal of the American Medical Association in September 2026 argues that artificial intelligence will soon surpass human physicians in patient care, suggesting that “AI-alone” care could outperform human-AI hybrid models [1]. Authors Ezekiel Emanuel, a bioethicist, and Vinod Khosla, a venture capitalist, contend that in cognitive medical functions, standalone AI medical care is likely to be better than physician-only or physician-AI hybrid care [1]. They predict that within four years, AI advancements will be significant enough to question the necessity of human oversight in diagnostics [1].
The American Medical Association (AMA) has issued a sharp response to these assertions, emphasizing critical limitations in autonomous diagnostic testing [1]. John Whyte, CEO of the American Medical Association, stated that while the AMA sees potential in these tools, they must be utilized in the context of a care plan that’s governed by a physician [1]. The organization highlights that clinical context, human intuition, and liability remain key hurdles that simulations alone cannot address [1].
Adoption Rates and Clinical Reality
Internal data from OpenEvidence, an AI chatbot for clinicians, indicates that approximately two-thirds of physicians in the United States currently utilize the tool, highlighting widespread adoption in clinical practice [1]. This rapid integration has occurred in the 46 months since the November 2022 release of ChatGPT, marking a period of accelerated technological advancement [1]. However, the reliance on internal company data for adoption metrics introduces variables that require independent verification [alert! ‘internal data not independently verified’] [1].
Recent events underscore the complexity of integrating AI into high-stakes environments. A team of neurosurgeons in the UK performed a brain tumor removal surgery guided by an AI tool, as announced on 2026-08-28, demonstrating practical application [1]. Conversely, credibility challenges persist; Springer Nature retracted a study that claimed ChatGPT had a “large positive impact on improving learning performance” due to major analytical discrepancies [1]. Robert Wachter, Head of medicine at the University of California, San Francisco, warned of the “doorman fallacy,” noting there will be times when humans will muck up the performance, but oversight remains crucial [1].
Regulatory Boundaries Emerge
As the technology evolves, federal and state regulators are drawing new boundaries around medical AI, from how the tools are vetted to who bears responsibility when they fail [2]. This regulatory scrutiny aligns with the AMA’s call for strict oversight and human-in-the-loop validation amidst the push for rapid market integration by technology developers [1]. The focus remains on ensuring accuracy and responsibility to avoid misinterpretation of AI-driven diagnostics [2][3].
The tension between innovation and safety continues to shape the landscape of healthcare enterprise adoption [1]. While developers push for autonomous capabilities, professional regulatory bodies advocate for frameworks that prioritize patient safety and clear liability structures [2]. The outcome of this discourse will likely define the operational parameters for AI in medicine for the foreseeable future [3].