Security & data handling
Read-only to start.
Your data stays yours.
Claims data is regulated, personal, and yours. The deployment posture is built around that — starting read-only, on your terms, with every AI action on the record.

01Data handling
What we read, where it lives, what we never keep.
NDA first
Every engagement — including the free audit — starts under NDA, before any file moves.
Read-only start
The audit step runs on a data export or read-only access. Nothing writes to your system until you approve automations — and then every write is approval-gated.
No-retention option
For the free audit, data handling is scoped to the engagement: findings delivered, source data deleted on request.
Demo data is synthetic
Everything on this site and in our demos uses illustrative, synthetic claims — labeled as such. Real data only ever moves under contract.
Access controls
Least-privilege access, scoped credentials per engagement, and audit logging on our side as well as the product's.
Compliance
SOC 2 program in progress; compliance documentation and subprocessor list available under NDA. We will never show a badge we haven't earned.
02AI governance
Every action, on the record.
- Every AI action logged in the claim file with its reasoning
- Sources cited on every finding — statutes, fee schedules, your SLAs
- Your adjusters make every call; nothing consequential is auto-acted
- Token and compute cost visible to the people who own the budget
- Your playbook, encoded — carrier rules, not a vendor template
HELD FOR REVIEW — sources in the margin, ambiguity waits for a person, the adjuster signs. Illustrative document, synthetic data.
03Connection modes
Meets your system where it is.
API-first where an API exists; browser-level where one doesn't. Five connection modes from read-only export to approval-gated write-back — an integration program is never the precondition.
- 01 Scope
- Least-privilege credentials
- 02 Observe
- Read-only by default
- 03 Approve
- A person gates every write
- 04 Record
- Every action stays on the file

Diligence
Security questions? Ask them early.
We'd rather walk your IT and compliance teams through the posture before the pilot than after.
Prefer email? audit@hypermodel.ai
