Your data needs visible boundaries
Designing a local workflow and explaining when a task uses an external service.
Local is a configuration, not a slogan
RadMesh is designed around an institutional node. Storing images and reports there does not mean every AI task runs there. Each deployment must explain which service processes each task and what information it receives.
Every transfer needs an explanation
Authorization, data minimization and an outgoing-transfer log are part of the design. When a feature uses an external service, its scope must be inspectable. Switching models should not hide a change in that boundary.
The website has a boundary too
Ask ChatGPT and Ask Claude share an editorial question and a public document URL. They do not attach studies, clinical conversations or private source code. Visitors can read and copy the question before opening the service.
What this note does not certify
This explains product architecture and communication. It does not certify regulatory compliance, installation security or suitability for real patient data. Those require deployment-specific verification.
Explore with your AI
An external service will open with this question and a link to our public overview. No clinical data or private code is attached. That service generates the answer; check its claims against the source.
View and copy the question
Read this RadMesh note: https://www.radicalmesh.com/about/notes-data-and-control-en.txt Explain “Your data needs visible boundaries”. Separate design principles, available features and plans. Cite the source. If you cannot open it, say so. Do not invent results or clinical validation.