MDCalc launches a tool to check doctors’ AI questions for privacy risks
CrossCover is designed to remove identifying information and check chatbot answers. MDCalc sees those safeguards as a way to encourage clinical AI use.
Loading page…
CrossCover is designed to remove identifying information and check chatbot answers. MDCalc sees those safeguards as a way to encourage clinical AI use.
Listen to this story
MDCalc launched CrossCover on November 11, 2026, with safeguards aimed at making clinical generative AI easier for physicians to use. It is designed to remove identifying patient and provider information from clinicians’ queries and separately check chatbot responses for accuracy, without requiring clinicians to submit protected health information under privacy agreements. These are stated functions, not guarantees that every identifier will be removed or every answer will be correct; the product’s practical value will depend on how reliably the checks work in use.
STAT contrasts CrossCover’s approach with services that require national-ID registration or agreements permitting clinicians to enter protected health information.
STAT says OpenEvidence, Abridge and Epic draw on expanding patient, provider and payer information for recommendations; the comparison does not establish an accuracy advantage.
MDCalc’s head of AI, Graham Walker, described CrossCover as an accelerator for clinical generative AI adoption.
MDCalc launched CrossCover on November 11, 2026, giving clinicians a tool designed to check both sides of a medical AI conversation: their questions for privacy concerns and the chatbot’s answers for accuracy. The medical reference company wants clinicians to feel more confident using these tools while preserving patient and provider privacy, according to STAT’s account of the announcement.
The privacy mechanism centers on what clinicians put into a query. MDCalc says CrossCover strips out identifying patient and provider information. Its answer-checking function addresses a separate concern: the accuracy of what the chatbot returns. These are the tool’s stated functions, not a guarantee that every identifying detail will be removed or every answer will be correct.
STAT contrasts that design with two other approaches to access and privacy: requiring clinicians to register using national identification numbers, and brokering privacy agreements that allow them to enter protected patient health information. Those agreements permit sensitive information to be included; CrossCover’s approach focuses on removing it from clinicians’ questions.
The comparison extends to the information used for medical recommendations. STAT describes OpenEvidence, Abridge and electronic health record company Epic as incorporating an expanding range of patient, provider and payer information into their recommendations. That is a comparison of product approaches, rather than evidence that one delivers more accurate answers.
“We can take away some of the challenges and risks and problems with the technology so that physicians can feel more confident in using it,”
Graham Walker, head of AI at MDCalc, speaking to STAT
Walker said MDCalc sees CrossCover as an accelerator for clinical generative AI. The company is positioning the checks as a way to support adoption, not as a reason for physicians to hold back from using the technology.
Loading discussion...
Join the conversation
Explain whether another safeguard would reassure you or leave your concerns unchanged.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.