Identifying data
Nearly 3,200 patients' data in an unapproved AI tool
In September 2026 an employee at Johns Hopkins Bayview in Baltimore uploaded a spreadsheet of patient data to a public AI tool the hospital had not approved. Patients were notified in October.
According to Johns Hopkins Bayview Medical Center, on 21 September 2026 an employee uploaded a spreadsheet to a public artificial intelligence tool that the hospital had not approved for use. It held data on nearly 3,200 patients who had visited the hospital between May and September.
The uploaded data included provider information, treatments and medications, as well as account numbers and medical record numbers. According to the hospital, the spreadsheet contained no Social Security numbers or financial information. The hospital did not name the tool.
When the incident was discovered, the hospital opened an investigation. The employee had to delete the original prompt, the thread and the uploaded document from the tool's history, and the hospital also asked the company operating the tool to delete the data. It did not say whether that happened. Affected patients received letters in October and were offered a year of free credit monitoring. The hospital said it had put further measures in place and is reinforcing its policies and staff training on handling health information.
What can we learn from it?
The case shows that once something is uploaded, deleting it can only be requested: the hospital could not say whether the provider had actually removed the data. AI can genuinely help with spreadsheet work; the question is with what data. Confiella sends only reviewed, pseudonymised text to the cloud, to an EU regional endpoint. With any other tool the same practice works: let the organisation name the AI tools that may be used, and give them only the part a task needs, without names and identifiers.
Sources
This article is based on the public sources listed. If you find a mistake, write to us: [email protected]