As Healthcare Breaches Double, ChatGPT Health Raises Security and Safety Concerns

A new report from Fortified Health Security says there were twice as many healthcare sector breaches in 2025 as there were in 2024. Ransomware and vendor breaches are fueling the rise. While the number of records exposed is down due to a lack of the massive (record count) breaches that we had in 2024, the number of breaches is now constant.

On the issue of vendor risk, only 4 percent of healthcare organizations had a high confidence that their vendor risk assessments were adequate. Almost a third said that they were not confident at all.

6 percent said they were very confident that they could quickly identify, contain and recover from an incident. In significant part, this lack of confidence is a resource issue – money and people. Other factors can be overcome, but not without significant additional resources. Credit: Cybersecurity Dive

With that as a background, ChatGPT just released their Health product. It is a large language model chatbot that will allow a consumer to give ChatGPT access to all of their health information.

Given AI’s reputation these days, ChatGPT is going out of their way to say that they intend for ChatGPT Health to be a more secure experience. It comes with additional layered protections.

While I am sure that it does, my level of confidence that they can out think the hackers is low, at best. OpenAI, trying to avoid being sued (as much) stresses that this should not replace medical care. But, likely, it will in many cases.

ChatGPT is “partnering” with health data firm b.well to directly hoover up your health data from doctors and hospitals that have APIs built on the Fast Healthcare Interoperability Resources (FHIR) as well as healthcare exchange frameworks.

While OpenAI says you can disconnect at any time, removing data that you have already shared is, at this point in time, impossible, so users should understand what they are doing prior to giving any AI permission to vacuum up your personal healthcare data. Credit: Data Reading

by