Paper

Zero-egress health AI with a patient-held consent layer

The design and early operation of the LifeCare network, written up in one place: the consent layer that holds no health data, the provider-operated nodes that keep raw values where they are, and the open-weight models that run beside them. Published on Zenodo with a permanent DOI.

Preprint, 20 September 2026. Raphael Gernot, Enbarca Inc. Licensed CC BY-NC-SA 4.0.

Read the paper

• DOI 10.5281/zenodo.22855810, the permanent identifier, which resolves to the record whatever else moves.
• The Zenodo record, with the metadata, the version history and the export formats.
• The PDF, the full text as published.
• CC BY-NC-SA 4.0, the licence: share and adapt with attribution, non-commercially, under the same terms.

What the paper describes

The paper sets out, in one document, the parts of the network this site documents a page at a time. It specifies the consent receipt issued for every grant and approved by a passkey assertion over the canonical grant, the patient-held authorization server that turns a grant into a purpose-bound, minimum-scope token and revokes it in under a second, the three layers of zero-egress enforcement that make "raw data never leaves" a property of code paths rather than policy, and the software-rooted attestation whose measurement anyone can recompute from a node's public response. It adds what the reference pages do not carry: the design rationale, erasure by key destruction that keeps the ledger verifiable, an account of how the first production node has been operated, the limits it states plainly, and where the design sits against personal data stores, patient-controlled authorization servers, confidential cloud inference and agent-memory layers.

Abstract

Personal health data is fragmented across device vendors, laboratories, clinics and apps, each holding it under its own terms, while a new class of AI assistants seeks to become the durable holder of an individual's longitudinal health context. We describe the LifeCare Network, a deployed system that separates three concerns the industry usually collapses into one: control of health data (a patient-held identity, consent and authorization layer that stores no health data), holding of the data (provider-operated nodes that keep raw values in place), and inference over it (open-weight models that run on the node, with only interpreted outputs leaving). We present the consent receipt shaped to ISO/IEC TS 27560 and approved by a passkey assertion over the canonical grant, a UMA-pattern authorization server with purpose-bound tokens validated by introspection and revoked in under a second, a hash-chained per-person ledger and resource registry, three enforcement layers that make "raw data never leaves" a property of code paths rather than policy, erasure by key destruction that preserves the ledger's verifiability, and a software-rooted attestation whose measurement anyone can recompute from the node's public response. We report early operating results from a single production node: a champion/challenger gate that rejected a stronger benchmark model on the network's own golden set, a conversational layer redesigned after prompt-only rules failed in live traffic, and a pre-registered twelve-week outcome study with locked endpoints and kill criteria. We state the limits: attestation is not yet hardware-rooted, enforcement of caller authentication is a deployment setting, and no outcome data exists yet. We position the design against personal data stores, patient-controlled authorization servers, confidential cloud inference, and agent-memory layers, and argue that the unoccupied position is data-scoped, purpose-bound consent enforced next to the data.

How to cite it

Gernot, R. (2026). Zero-Egress Health AI with a Patient-Held Consent Layer: Design and Early Operation of the LifeCare Network. Zenodo. https://doi.org/10.5281/zenodo.22855810

@misc{gernot2026lifecare,
  author    = {Gernot, Raphael},
  title     = {Zero-Egress Health AI with a Patient-Held Consent Layer:
               Design and Early Operation of the LifeCare Network},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.22855810},
  url       = {https://doi.org/10.5281/zenodo.22855810},
  note      = {Preprint}
}

A preprint, not a peer-reviewed paper. This work has been posted to a public repository and given a permanent identifier. It has not been through journal peer review, and no claim on this page should be read as carrying that endorsement. It describes a general wellness system and is not clinical guidance. The DOI above is version-independent, so it keeps resolving if a later version is posted.