Each year, medical diagnosis errors affect the health of millions of Americans and cost billions of dollars. Machine learning technologies can help identify hidden or complex patterns in diagnostic data to detect diseases earlier and improve treatments. We identified such technologies in use and development, including some that improve their own accuracy by learning from new data. But developing and adopting these technologies has challenges, such as the need to demonstrate real-world performance in diverse clinical settings. Our policy options—like improving data access and collaboration—may help address the challenges.