The scan looked normal. More than two years later, a large pancreatic tumor was visible. In a case illustrated in a 2026 study, artificial intelligence identified suspicious patterns in the earlier image when researchers analyzed it retrospectively. That lost window is where AI’s promise becomes personal. Researchers are learning to detect warning signs before a tumor becomes visible, while a major trial has already demonstrated better breast cancer detection. With more than 2.1 million new cancer cases and 626,140 deaths projected in the United States this year, these advances deserve public attention. America’s response should make it easier to bring useful innovations to patients, with evidence guiding adoption and people retaining control over their care.
Consider what happened when researchers gave radiologists another way to examine mammograms. Sweden’s MASAI randomized trial involved roughly 106,000 women, comparing AI-supported screening with conventional readings by two radiologists. Earlier results found 338 cancers in the AI group versus 262 in the comparison group, an increase of 29 percent. The final results published in January 2026 strengthened the case for the approach. Screening identified 81 percent of cancers versus 74 percent, while false-positive rates remained similar. Radiologists also performed 44 percent fewer screen readings. That combination matters: more cancers found, without a comparable increase in false alarms, using less specialist reading time. It gives doctors a practical reason to consider the technology and patients a concrete reason to welcome its careful use.
Pancreatic cancer presents a harder challenge because early changes can escape conventional imaging. Mayo Clinic’s REDMOD system searches tissue texture and structure for patterns that suggest disease. In an independent test involving 493 patients, it identified 73 percent of prediagnostic cancers, compared with about 39 percent for radiologists, with a median lead time of 475 days before clinical diagnosis. Its intended use is especially appealing: examining CT scans already obtained for other reasons, particularly among patients at elevated risk. These were retrospective findings, and prospective testing remains necessary before broad clinical adoption. Still, finding additional value in an existing scan could make earlier detection more practical. The opportunity is to give a physician useful information while there is still time to investigate it.
AI’s usefulness may extend beyond the first diagnosis. On September 10, 2026, Mayo Clinic announced research examining whether routine pathology slides could reveal which pancreatic cancers were more likely to return. The study in Clinical Cancer Research analyzed tissue from 203 patients whose tumors had responded only modestly to treatment before surgery. Researchers used AI to map how cancer and surrounding tissue were arranged; more fragmented, intermixed patterns were associated with earlier recurrence. The approach could extract extra information from slides already produced during routine care. Researchers explicitly say prospective confirmation is needed before it guides clinical decisions. The prospect is compelling nonetheless: a familiar laboratory sample could help a doctor better understand an individual patient’s risk and, eventually, tailor follow-up care more precisely.
Results like these make a strong case for freedom to innovate with accountability. The FDA already reviews AI-enabled medical devices through established pathways, and HIPAA protects health information held by covered entities and their business associates. Americans for Prosperity’s AI policy framework sensibly asks policymakers to identify a concrete problem and determine whether existing law addresses it before adding another rule. Developers should demonstrate performance for the intended clinical use, clinicians should understand limitations, and patients should receive clear information about their privacy rights. Where a genuine gap exists, the response should be specific and enforceable. Those expectations give innovation credibility. They also make it possible to distinguish a promising experimental tool from a product ready for clinical use, without treating every application of AI as the same problem.
Getting good tools into ordinary clinics also requires room for competitors to build and improve them. AFP’s case for open experimentation and its AI policy reform agenda offer a useful starting point: avoid restrictions that needlessly burden new entrants or fragment the market. That risk is already materializing. States including Ohio, Maine, and Delaware have proposed or enacted laws that would bar AI from diagnostic and clinical roles altogether, and a growing patchwork of conflicting state disclosure and prior-authorization mandates threatens to slow the path from research breakthrough to the exam room. Private investment and research commercialization help turn discoveries into products that physicians can actually use. The computing infrastructure behind those products also needs abundant, reliable electricity, making permitting reform relevant even to the examination room. AFP rightly argues that data centers should bear their own energy and connection costs. Removing barriers to construction can support innovation while protecting households from having those costs shifted onto their bills. That is consistent with empowering entrepreneurs to solve problems and take responsibility for the resources they require.
The measure of success should be what patients gain. Better screening, earlier warning signals and more informative pathology are tangible reasons for optimism, with different levels of evidence behind each. Researchers now have more ways to investigate the years before disease becomes apparent. Public policy should preserve patient choice and the freedom to test new ideas as that work advances. For someone whose scan contains a clue that medicine cannot yet reliably recognize, progress could mean an earlier conversation with a doctor and another chance to act. We should give that possibility every reasonable opportunity to become part of everyday care.
Hayley Bieron is a Policy Associate at Americans for Prosperity.
