05/06/2025
The dangers of AI in healthcare.
With the introduction of easily accessible public AI. A few medical practices have discovered the hard way, that you need to. 1, understand the technology. 2, have and distribute a policy for the business.
Take this example. A healthcare worker would like a pie chart of all patients that are medium to high risk in their area, area X. For an educational project they are working on.
They download a list of all patients in those ranges. Then they ask Chat GPT to produce a pie chart. Chat GPT does create a nice pie chart.
The problem is that the user has unknowingly given that patient information to the world! After this upload, anyone that say, asks chat GPT, give me a list of people in area X that are at risk of diabetes. They will receive a report with all of the information uploaded. Name, address, date of birth! Anything that was in that initial upload.
The user didn't intend to share this information, neither did the business. But both did not understand the risk.
When using public AI. the information is shared publicly. Anything you upload to is it used to learn.
You can use private AI. Microsoft have a solution. In Healthcare, Heidi is an excellent solution. These products contain the data.
At the very least, we recommend all business should have and communicate an AI policy. Without that, you can forgive a young worker for looking for shortcuts. When there is a policy, this is less of a risk.
The policy should contain a method for marking that the data produced was generated using AI.
Also, when using AI with an individual. You need to have a consent process that is logged in your system. Some AI apps will log it automatically if you ask the other person if they consent to using AI to take notes.
I have written this raw. (no AI). So it may be less elegant, but I think it's best when writing about AI to not use AI :)
Call now to connect with business.
