MDCalc has launched CrossCover, a tool that screens clinicians’ questions before they reach an AI model and checks some answers against medical references. It does not write answers itself, and its answer-checking feature is still in beta, so the useful question is what protection it offers in practice, not whether it makes clinical AI risk-free.
Watch Desk analysis
What happened
MedicalDaily reports that CrossCover launched on 11 November. Its four features include detecting likely patient identifiers and recommending their removal; limiting details about the clinician, such as identity and location, from reaching partner models; comparing AI-generated clinical information with references from ECRI and the American College of Physicians through the beta feature CrossCheck; and surfacing a relevant MDCalc clinical calculator when a question calls for one.
Partner services, initially AMBOSS and Vera Health, generate the answers. CrossCover applies its checks around them. MedicalDaily says MDCalc is opening access to a founding group of practising physicians and expects to add more AI partners.
Why it matters
Clinicians may use chatbots for quick clinical look-ups, where a fluent but outdated or incorrect answer can matter. Screening for identifiers and flagging differences from references could add useful checks to that workflow. The distinction matters: CrossCover does not replace the partner models or make their answers automatically reliable.
Our read
This is a practical attempt to put guardrails around existing clinical chatbot use, rather than to sell another chatbot as the doctor in the room. But MedicalDaily says no independent evaluation of CrossCover has been published, and CrossCheck remains in beta. A reference flag is a prompt for review, not a verdict. Patients can ask whether their clinician uses AI, whether a person reviews its output and how their information is protected.
What to watch
- Whether independent testing shows how often CrossCheck catches errors, or misses them.
- How CrossCover handles patient details that could identify someone without a name or record number.
- Whether more AI partners join and which references their answers are checked against.
Discussion spark: Would you trust a clinician to use an AI chatbot with safeguards like CrossCover, or should clinical AI use wait for independent testing?
Sources and evidence
- Source update (13 November 2026, 14:17 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.