FDA plans final AI device, surgical robot guidance; five systems hold off on Epic AI tools; Senate passes hospital cyber bill; clinicians favor AI discharge notes

A pinboard in a navy frame with two white cards, one showing a microchip and one a robotic arm, pinned at the top with glowing amber pushpins.

The Food and Drug Administration plans to finalize guidance on AI-enabled devices and surgical robots in fiscal 2027 and to draft evidence recommendations for generative AI conversational devices for mental disorders, its priority list shows.

The device center's list for fiscal 2027, which MedTech Dive reported Oct. 2, puts seven final guidance documents and one draft on its top tier. Separately, five health systems told Becker's Hospital Review they are holding off on some of Epic's AI tools, and Children's Healthcare of Atlanta said one of them produced too many inaccuracies about diagnoses and patient locations.

The Senate passed a hospital cybersecurity bill with provisions for rural providers by unanimous consent Oct. 1, sending it to the House. In a Stony Brook study, clinicians preferred discharge summaries drafted by a large language model over human-written ones in 95% of 60 longer hospitalizations.

  • Regulation Oct 2, 2026
    The FDA's Center for Devices and Radiological Health lists seven final guidance documents it intends to publish in fiscal 2027, including "Marketing Submission and Lifecycle Management Recommendations for Artificial Intelligence-Enabled Device Software Functions," and premarket submission recommendations for robotically assisted surgical devices. The only draft on the A-list is "Generative Artificial Intelligence-Enabled Conversational Devices for Mental Disorders: Evidentiary Recommendations." The surgical robot draft was posted in September and is open for comment until Nov. 24, MedTech Dive reported. The second-tier B-list includes a draft policy on device software functions, a draft on postmarket cybersecurity management and final guidance on pulse oximeter testing and labeling.
  • Deployment Oct 2, 2026
    Children's Healthcare of Atlanta put Epic's inpatient insights feature on hold because it produced too many inaccuracies about diagnoses and patient locations and overly long narratives, Chief Information Officer Jeremy Meller told Becker's Hospital Review, adding that the product is on a cycle for reevaluation. St. Luke's Health System in Boise, Idaho, is staying with Ambience Healthcare for ambient documentation rather than Epic's Art, and Penn Medicine is holding off on Epic's Emmie, its ambient documentation and chart summarization while it relies on Ambience and its own tools. Atlantic Health in New Jersey is taking "a measured approach to tools that make autonomous clinical decisions or engage patients without appropriate human oversight," Chief Information Officer Sunil Dadlani said. Baptist Health in Jacksonville, Florida, said it is weighing maturity and return on investment.
  • Evidence Oct 2, 2026
    Hospitalists and primary care physicians compared summaries written by a large language model with human-written ones for 60 internal medicine patients hospitalized seven to 21 days at Stony Brook Medicine, the authors reported in npj Digital Medicine on Oct. 2. Reviewers preferred the model's version in 95% of cases, rated it higher for quality, readability, factuality and completeness, and annotated fewer factual errors or clinically relevant omissions in it. The authors found no significant difference in estimated potential for harm; primary care physicians flagged more omissions and assigned higher harm likelihood than hospitalists. Of 31 incidental radiology findings in the model's summaries, 93.5% were stated correctly and 87.1% were judged appropriate to report. The evaluation was retrospective.
  • Evidence Oct 3, 2026
    Researchers from Seoul National University, Kookmin University and Seoul National University Hospital tested B-RAD, an interpretable retrieval-augmented system for diagnosing breast ultrasound, on 8,311 images from 11 cohorts in seven countries, according to a study published Oct. 3 in npj Digital Medicine. Without further tuning it scored an area under the curve of 0.952 for deciding which lesions need biopsy on an independent institutional cohort, the authors said. In a reader study of four radiologists, the system's help improved accuracy, raised agreement between readers from moderate to substantial and reduced missed malignancies for every reader. The validation was retrospective, and the reader study was small.
  • Regulation Oct 1, 2026
    The Senate passed the bill from Sen. Bill Cassidy, a Louisiana Republican and physician, on Oct. 1, according to the Senate health committee. It is meant to strengthen coordination among federal agencies on health care cybersecurity and improve security practices at rural providers; the release does not state an authorized dollar amount. Co-sponsors include Sens. Maggie Hassan of New Hampshire, John Cornyn of Texas, Mark Warner of Virginia, Cindy Hyde-Smith of Mississippi and Angus King of Maine. Hassan said the bill would help facilities with fewer resources "strengthen cybersecurity and respond faster to attacks," and Warner urged the House to act, Becker's Hospital Review reported.
  • Evidence Oct 1, 2026
    University of Vermont researchers reported in JAMA Health Forum that 10.5 million of the nearly 12 million people in areas below 100 megabits per second download and 20 upload speeds live in rural areas, Healthcare IT News reported Oct. 1. About 650,000 rural residents face gaps in broadband, ambulance service and health care facilities at once, and 946,000 rural residents have adequate broadband but no nearby ambulance or health services. The South accounts for 6.5 million of those affected, and 32% of rural residents in the West live in broadband deserts. The authors called for targeted infrastructure investment and expanded reimbursement so telehealth can reduce gaps in access.
  • Incident Sept 30, 2026
    The model "uncovered a critical vulnerability exposing sensitive personal information across healthcare software used by hospitals worldwide," a risk earlier frontier models had missed, Google said Sept. 30. The security firm Wiz, which Google owns, ran the search through its Scan for Good program, which looks for high-risk exposures in critical public infrastructure for free. Neither the software vendor nor the hospitals were named, and Becker's Hospital Review reported Oct. 2 that it was not stated whether the flaw has been patched.