How VetIOS handles clinical data, inference traces, and operator telemetry across the platform mesh.
VetIOS is designed with a privacy-first clinical mesh. We collect three primary categories of data to facilitate veterinary intelligence:
Traffic is encrypted in transit with TLS 1.2 or later. The database is encrypted at rest by the hosting provider (Supabase). Row-level security scopes clinical rows to their tenant.
VetIOS does not sell clinical data. Clinic-submitted data is not used to train model weights, and no pipeline reads tenant databases into training data. Outcome diagnostics are tenant-scoped and preserve the original inference scores. Displayed scores are not adjusted by legacy calibration tuples. Supabase (database, authentication, storage) and Vercel (application hosting) process all platform data; see data residency. Stripe processes billing. Clinical content leaves VetIOS infrastructure only for AI features: images in Ask VetIOS image analysis go to Anthropic, and images for vision inference and assistant text go to an OpenAI-compatible provider. The structured differential engine makes no external model call.
VetIOS is the controller for VetIOS accounts, login and security telemetry, billing, fraud and security records, website analytics, direct support, and direct PetPass consumer accounts. For identifiable owner and patient information a clinic submits, the clinic is the controller and VetIOS processes it only on the clinic's instructions. These positions are pending counsel review.
Clinic-submitted data enters shared validation, research, federation, or population intelligence only in de-identified or aggregate form and only with the clinic's matching consent: deidentified_training for shared validation and research, network_learning for federation, population_signal for population intelligence. Direct patient and owner identifiers never do.
Data at rest: Supabase (Postgres, Auth, Storage) in AWS Europe (Ireland), eu-west-1 (observed 2026-09-15). Application processing: Vercel (serverless functions for the web app and API) in Vercel iad1, Washington, D.C., USA (AWS us-east-1) (observed 2026-09-26). Requests are processed in United States and data is stored in Ireland (EU), so clinical data submitted to VetIOS crosses between those jurisdictions. Content sent to external AI model providers is processed on their infrastructure, whose location VetIOS has not verified.
This is the canonical data-use model. The support page and trust ledger render the same records.
Name, email, and authentication identity of clinic staff and operators who sign in to VetIOS.
Sign-in events, rate-limit and abuse signals, credential use, and security incident records.
Subscription, invoice, and payment status records. Card data is handled by the payment processor, not VetIOS.
Emails and messages sent directly to VetIOS support.
VetIOS runs no website analytics or tracking scripts today. If introduced, VetIOS is the controller.
An owner account created directly with PetPass, including notification preferences.
Signalment, presenting signs, history, exam findings, lab values, and images a clinic submits.
Patient names, microchip numbers, and similar direct identifiers.
Owner names and contact details held in a clinic's records.
The input signature, ranked differentials, CIRE block, and lineage recorded for each inference.
A confirmed diagnosis or outcome attached to a prior inference, with its evidence type.
Per-tenant calibration buckets: outcome counts, Brier score, and calibration error per label, species, and model version.
Content sent to a third-party AI model provider: images for vision analysis and text for the assistant.
Append-only records of governance gates, reliability packets, calibration snapshots, and review queues about clinic-submitted data.
A masked, de-identified contribution a consenting tenant publishes to a federation round.
Population-level counts and advisories with no clinic, patient, owner, or inference identifiers.
A de-identified dataset file prepared for research or shared validation.
vetios-data-governance-2026-09-26b
VetIOS uses Google OAuth through Supabase Auth for sign-in only. The integration requests standard OpenID Connect scopes and does not access Google Drive, Calendar, Gmail, Workspace APIs, or any other Google service. VetIOS requests minimum scopes only.
You can revoke VetIOS Google access from your Google Account permissions page at myaccount.google.com/permissions.
To request VetIOS account deletion, contact privacy@vetios.tech.
VetIOS does not use Google user data, including email, profile, or any Google-sourced information, to train, fine-tune, evaluate, or improve any AI or machine learning model. Any VetIOS training or evaluation workflow must use an explicitly approved, provenance-tracked dataset under the applicable consent and licensing controls. Google-sourced identity data is isolated from those workflows.
As a platform operator or partner, you retain ownership of your clinical inputs. Email privacy@vetios.tech to request export or deletion. The platform team handles requests, with a 30-day target, and records every action in an append-only erasure ledger. Data is hard-deleted where possible and anonymized where identity is unnecessary. Records that must be kept, such as the audit record of a published clinical decision, are restricted: the identity link is removed, the payload is minimized, the legal basis is recorded, and processing is limited to the recorded purposes. There is no automated retention or purge schedule today. Data is kept until a deletion request is executed. Learning uses are consent-scoped per tenant: deidentified_training (shared validation and research), network_learning (federation), and population_signal (population intelligence). Consent can be revoked; revocation does not unwind aggregates already produced.
For inquiries regarding GDPR, CCPA, or specific clinical data handling policies, please reach out to:
privacy@vetios.tech