Your clinicians are signing AI-drafted notes. Your malpractice coverage was written before they started.
Idem validates the clinical AI your hospital already runs, on your own patients, and ties the result to coverage your risk team can use.
The AI in your record was activated, not procured.
If your hospital runs Epic, Oracle Health, or Meditech, AI is likely already summarizing charts, drafting discharge notes, and pre-replying to in-basket messages. Those features were switched on during an upgrade. Nobody tested them on your patients before your physicians started signing their output.
When someone did test, the results were poor. In May 2026, Ontario's Auditor General reviewed all 20 AI scribes approved for clinical use in the province. Every one produced inaccuracies. Nine fabricated content. Twelve captured a different drug than the one prescribed.
You carry the risk, and nobody has priced it.
71% of US hospitals run predictive AI inside the EHR, and only 18% of health systems govern AI maturely. The Joint Commission and CHAI now expect local validation and ongoing monitoring, but nobody funds it. Your malpractice carrier covers AI by default and has no way to see what you run.
Validation that ends in coverage.
- 01 Inventory Every clinical AI tool in use, including EHR features on by default.
- 02 Local validation Performance on your patients and your workflow, against pass and fail criteria your clinical and risk leaders agree to before testing.
- 03 Signed finding Error rates by failure mode and patient group, reviewed and signed by a named clinician.
- 04 Coverage A tier your carrier can act on. We are building coverage and premium credit programs with underwriting partners so a passing validation earns better terms.
The engagement letter is signed by your organization, not the vendor, and the findings are yours.
Coverage built for how clinical AI fails.
Fabricated content that ends up in the medical record, like a symptom the patient never reported or a history that never happened.
A different drug than the one prescribed, a missed finding, a dropped allergy, or a summary that leaves out what mattered.
Performance that changes after a vendor update or as the patient mix shifts, caught by revalidation every policy term.
Error rates that climb for specific patient groups the vendor never tested on.
Final terms, limits, and exclusions are set in the policy form with our underwriting partners.
CMIOs, chief AI officers, CMOs, chief risk officers, risk managers, and captive managers at hospitals and health systems. Best fit when at least one EHR-native AI feature or scribe has been live for six months or more, and a malpractice renewal, board question, or accreditation survey is coming up.
Read the evidence.
Then let's talk.
The June 2026 white paper documents what EHR-native AI features are doing in the medical record and what your organization is on the hook for.