One Spreadsheet, One Unapproved AI Tool, 3,182 Patients: The Johns Hopkins Bayview Exposure at a Glance
No hacker was involved. A Johns Hopkins Bayview Medical Center employee uploaded a spreadsheet of patient information to a public AI tool the hospital had not approved. Here is the incident fact file and what it teaches any business whose staff use AI.
Most breach notices start with an intruder. This one starts with an ordinary task at a desk. A worker at Johns Hopkins Bayview Medical Center in Baltimore uploaded a spreadsheet of patient details to an outside AI tool that the hospital had not approved for that use, and more than 3,000 patients are now receiving letters about it.
The Baltimore Banner reported the incident on Sunday, based on a Johns Hopkins release, and WMAR-2 News published additional details from the hospital’s notice.
Fact file
- Organization: Johns Hopkins Bayview Medical Center, Baltimore, Maryland.
- What happened: An employee mistakenly uploaded a spreadsheet with patient information to an unnamed third-party AI platform that was not approved for that purpose.
- When: Late September. WMAR reports the hospital discovered the problem on September 21, the same day the upload happened, during an internal review.
- Who is affected: 3,182 patients who were seen at the center between May and September 2026.
- Data involved: Patient names, admission and treatment dates, medical record numbers, account numbers, the treating department, medications and provider names.
- Data not involved: Social Security numbers and financial account information, according to the hospital.
- Containment: The employee deleted the prompt, the conversation and the uploaded document from the tool, and Johns Hopkins asked the company behind the tool to delete the data.
- Notification: Letters began going out on October 22 and offer 12 months of free credit monitoring through Experian IdentityWorks. The hospital has set up a toll-free call center at 1-833-931-7766.
- Follow-up: The hospital says it is strengthening policy and training on approved and unapproved apps and AI tools. It is not clear whether the employee will face consequences.
Why this one is different
There was no malware, no stolen password and no ransom note. The data left because a well-meaning person used a convenient tool that was not on the approved list, sometimes called shadow AI. The hospital caught it the same day, which is better than many organizations would manage, but once information is pasted into an outside service the business depends on that provider to delete it. The hospital has asked; the reports do not say whether the provider has confirmed deletion.
Lessons any business can use
- Decide which AI tools are approved, write it down in one page, and tell staff what kinds of data may never go into any of them, such as customer records, health details, payroll or contracts.
- Give people an approved option. Staff turn to unapproved tools when they have a real task and no sanctioned way to do it.
- Use the controls you already have. Many web filters, data loss prevention tools and browser management settings can warn on or block uploads to unapproved AI sites.
- Practice the response. Know who contacts the AI provider, how deletion is requested and documented, and when legal and notification duties kick in.
- Train with real examples like this one. A short story about a spreadsheet sticks better than a policy paragraph.
What this means for your company
If your team uses AI tools, and most do whether or not anyone has approved them, the biggest risk may not be an attacker at all. It is a busy employee pasting sensitive data into the wrong box. You do not need a large program to reduce that risk: an approved tool list, a clear rule on what data stays out, some basic technical guardrails and a plan for the day it goes wrong will cover most small businesses, whatever AI products you use.
Need help writing a practical AI use policy and putting guardrails around sensitive data? Reach out to Layer 7 Data Solutions to discuss how we’ve helped companies similar to yours.
About The Author
