Yes, and that’s actually the direction many healthcare organizations are starting to explore.
One option is exactly what you’re describing: ring-fencing the data so it never leaves a controlled environment. That can be done by running AI models inside a hospital’s private infrastructure or through enterprise agreements where the data stays within a protected security boundary.
The challenge, though, is that many clinicians are already using external AI tools in their day-to-day workflow because they’re fast and easy. Once notes are copied into those tools, the ring-fence is broken.
That’s why solutions like pre-processing or semantic masking (the “smart buffer” idea) can help. They remove patient identifiers before the text reaches the model, so clinicians can still use AI while the sensitive data stays protected.
Sure my there is a way to ringfecnce the data by security to keep the data safe?
Yes, and that’s actually the direction many healthcare organizations are starting to explore.
One option is exactly what you’re describing: ring-fencing the data so it never leaves a controlled environment. That can be done by running AI models inside a hospital’s private infrastructure or through enterprise agreements where the data stays within a protected security boundary.
The challenge, though, is that many clinicians are already using external AI tools in their day-to-day workflow because they’re fast and easy. Once notes are copied into those tools, the ring-fence is broken.
That’s why solutions like pre-processing or semantic masking (the “smart buffer” idea) can help. They remove patient identifiers before the text reaches the model, so clinicians can still use AI while the sensitive data stays protected.