CARE

Research

Working paper

Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation

A five-class scheme for labelling how sensitive a request is, and routing it only to the set of models allowed to see it.

Status
Working paper
Published
Author
Michael Turon
Full text
Pending
Cite
Jump to citation
On this page

In short

Link
Thesis: Every query has a sensitivity profile. The right architecture classifies first, routes to an admissible model set second, and audits the routing decision for every call.

Key points

Link
  • Every query has a sensitivity profile, so the choice is not cloud or local: each request needs its own route.
  • Classify each request by how sensitive it is: health data, personal data, privileged, classified, or general.
  • Route it only to models cleared for its label, on a device or in the cloud, then pick the best one on cost and speed.
  • Audit every choice, so a reviewer can check what went where, and why.
  • The full text is pending. The flows on this site are illustrative, not measured data.

The five classes

Link
Health data
“Summarize this patient’s latest lab results.”Sent to On-device modelHeld from Frontier model
Personal data
No sample on this site yet.
Privileged
“Summarize this deposition.”Sent to On-premises model, Frontier model (split)Held from Frontier model (case facts)
Classified
“Help with this signal pattern.”Sent to Accredited system on the secure networkHeld from Frontier model
General
“Does Drug A clash with Drug B?”Sent to Frontier modelHeld from None

Samples: illustrative flow — not measured data.

What we don’t know yet

Link
  • How much time a classifier model adds when it runs on the same device. We have not measured that yet.

Cite

Link
Citation format 
Michael Turon (2026). Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation. Working paper. CARE Institute (Center of Agentic Research and Education). https://careinstitute.ai/research/privacy-routing/
@misc{turon2026beyond,
  author       = {Turon, Michael},
  title        = {{Beyond Cost-Quality: Privacy-Aware Routing for Local-to-Cloud LLM Escalation}},
  year         = {2026},
  note         = {Working paper},
  howpublished = {CARE Institute (Center of Agentic Research and Education)},
  url          = {https://careinstitute.ai/research/privacy-routing/}
}