A Johns Hopkins Bayview worker uploaded 3,182 patients' records to a public AI tool the hospital never approved

Johns Hopkins Bayview Medical Center in Baltimore is notifying 3,182 patients that their health information was exposed after a staff member uploaded a spreadsheet of patient data to a public AI tool the hospital had not approved for that use.

This is "shadow AI" with a patient count attached. Nobody hacked anything: one employee doing an ordinary task pasted a spreadsheet into the nearest chatbot, and 3,182 people's medications and record numbers left the building. The containment plan, delete the chat and ask the vendor nicely to delete its copy, shows how little control a hospital has once data hits a consumer AI tool; whether the data was logged, retained or used for training depends entirely on a company the hospital won't even name. Expect more of these as clinicians reach for AI to save time, and expect regulators to start asking hospitals not just which AI tools they approved, but how they stop staff from using the ones they didn't.

Johns Hopkins Bayview Medical Center says it discovered the problem on Sept. 21, 2026. That day, during an internal review, a staff member uploaded a spreadsheet of patient information to an outside, public AI platform. The hospital had not cleared that tool for this use. A public AI platform is a chatbot anyone can open, not a system the hospital runs. Those lines are the hospital's, as WMAR-2 News and The Baltimore Sun reported them.

The spreadsheet covered 3,182 patients admitted and treated between May 2026 and September 2026. That count and that window are the hospital's.

Exposed details may include patient names, dates of admission and treatment, medical record numbers, account numbers, the treating department, medications given, and the names of their providers. A medical record number is the ID the hospital uses for a patient's chart. The hospital says the file did not include Social Security numbers or financial account details. Those lines are the hospital's.

The hospital has not named the AI tool, or the company that runs it.

After learning of the upload, the hospital investigated. The worker deleted the AI prompt, the chat thread, and the document from the tool's history. The hospital also asked the company that runs the tool to delete the data. A prompt is what a person types or uploads to the chatbot. Those lines are the hospital's.

The hospital says it is tightening policies and training on protected health information, the patient data a hospital is required to keep private. That includes limits on unapproved apps and AI tools, and retraining staff on which tools are approved. Those lines are the hospital's.

Letters to the affected patients began going out Oct. 22, 2026. The letters offer 12 months of free Experian IdentityWorks credit monitoring. A toll-free line is open at 1-833-931-7766, engagement number B174716, Monday through Friday, 8 a.m. to 8 p.m. Central time. Those lines are the hospital's.

The picture is the Johns Hopkins Medicine Bayview Campus sign in front of the hospital buildings in Baltimore. The credit is a Baltimore Sun file photo, via Capital Gazette.

In plain terms, a Bayview staff member put a spreadsheet of 3,182 patients into a public AI tool the hospital had not approved. The file could include names, record numbers, medications, and doctors' names. The hospital says it did not include Social Security numbers or financial account details. The chat was deleted, the hospital asked the unnamed company to delete its copy, and letters started going out Oct. 22.