Stat Physician Physician Services
Services
AI Agents
Who we serve
Pricing
About
Book a consultation Call 248-853-9097
info@statphysician.com · Mon–Fri, 7am–7pm ET · after-hours line for active clients

AI trust & security

Built to be defensible, not just compliant.

HIPAA is the floor, not the ceiling. Below is exactly what we hold today, what is designed into the platform, and what we do not have yet — because a security page is only worth reading if it is true.

Plainly

What we do not have yet

We are an early-stage company. We do not hold a SOC 2 report, we are not HITRUST certified, and we have not completed a third-party penetration test. Plenty of vendors our size imply otherwise on a page like this. We would rather tell you plainly and let you judge the controls we can actually demonstrate today.

Every row in the status table below is reviewed at least quarterly. When our position changes, this page changes the same week — not when a deal depends on it.

The six rules that are not configurable

  • No agent ever diagnoses, triages, prescribes, or interprets a result.
  • No note, charge or claim is released without a human signature or a certified reviewer.
  • No PHI leaves the United States, and none of it trains a shared model.
  • No write-off, refund, disclosure or records release happens without human approval.
  • No password is shared, reused or stored in plain text — including payer portals.
  • No configuration change reaches production without a version, a test and a rollback path.

Controls

Six areas, documented and reviewable.

These describe how the platform is designed and what we commit to contractually. Where something is not yet built, tested or audited, the status table below says so rather than leaving you to guess.

Protected health information

  • A signed Business Associate Agreement before any PHI moves — ours or yours, your choice
  • PHI encrypted in transit (TLS 1.2+) and at rest (AES-256), with keys managed in a dedicated key service
  • US-only data residency, including backups and disaster recovery
  • Contractual prohibition on using your PHI to train any shared or third-party model
  • De-identification at the edge wherever a task does not require identified data
  • Retention periods set by you and enforced automatically, with deletion logged and provable

Access control

  • Named individual accounts only — no shared logins, ever
  • Multi-factor authentication required for every human with PHI access
  • Role-based least-privilege access, reviewed quarterly and revoked on offboarding within one hour
  • Agent service accounts scoped to specific systems, actions and time windows
  • Secrets and payer portal credentials held in a vault, never in code or configuration files
  • Break-glass access requires a reason, is time-boxed and triggers notification

Audit and accountability

  • Immutable audit log of every agent action: input, decision, action, confidence, model version, timestamp
  • Reasoning trace retained so any decision can be reconstructed months later
  • Human reviewer identity recorded on every approval, correction and override
  • Logs exportable as CSV or JSON at any time, including at contract end
  • Anomaly detection on access patterns and bulk record access
  • Annual third-party penetration test with a remediation report available under NDA

Agent guardrails

  • Confidence thresholds: below threshold, the agent cannot act — it queues for a human
  • Scope limits: each agent has an explicit list of systems and actions it may touch
  • Clinical boundary: no agent diagnoses, triages, prescribes or interprets results
  • Reversibility: bookings, filings and postings can be undone with a full trace
  • Kill switch per agent, per workflow and per practice, effective within minutes
  • Prompt-injection and jailbreak testing against every agent before each release

Regulatory position

  • HIPAA Privacy, Security and Breach Notification Rules — administrative, physical and technical safeguards documented
  • 21st Century Cures Act information-blocking rules respected in every integration
  • TCPA-compliant consent, opt-out and quiet-hours handling for SMS
  • State two-party consent rules applied to call recording by area code, with consent logged
  • Telehealth and documentation rules tracked state by state as they change
  • We are a business associate, not a covered entity — we never sell data, and we do not advertise to your patients

Operations and resilience

  • Encrypted, tested backups with documented recovery time and recovery point objectives
  • Documented incident response with breach notification inside the regulatory window
  • Failover: if an agent is unavailable, calls route to your existing line and work queues to your staff
  • Change control with versioned prompts and models, and rollback for every release
  • Workforce HIPAA training on hire and annually, with completion recorded
  • Subprocessors listed in writing, each under its own BAA, and you are notified before any change

Status

What is in place, today.

ControlStatusDetail
HIPAA safeguardsIn placeAdministrative, physical and technical safeguards documented, with a signed BAA before any PHI is exchanged.
Business Associate AgreementsAvailable nowWe sign ours or yours before a single record moves, and every subprocessor signs one too.
Data residencyUnited StatesWe do not process, store or back up protected health information offshore.
Model training on client dataNeverContractual, and it flows down to every model provider and subprocessor we use.
EncryptionTLS 1.2+ / AES-256In transit and at rest on every system we control, with managed key rotation.
Access controlMFA + least privilegeNamed individual accounts, no shared logins, quarterly access review, one-hour revocation on offboarding.
Audit loggingDesigned inImmutable action log with reasoning trace, model version and reviewer identity on every agent action.
SOC 2 Type IINot held — preparingWe do not have a SOC 2 report today and will not claim one. Ask us where our readiness work stands.
HITRUSTNot pursuedHITRUST is a later-stage certification for us. We will say that rather than imply otherwise.
Penetration testingNot yet completedWe will commission independent testing before enterprise deployment and share the report under NDA.

Rows marked Not held, Not pursued or Not yet completed are our honest gaps. We update them in tools/agents.py → SECURITY_STATUS the week our position changes, so this page stays ahead of the sales conversation rather than behind it.

FAQ

Security questions, answered plainly.

Are you HIPAA compliant?
We operate under HIPAA's requirements as a business associate: signed BAAs, documented administrative, physical and technical safeguards, workforce training, access controls, encryption, audit logging and breach procedures. We will give you our security packet and complete your security questionnaire before you sign anything.
Where is patient data stored?
In the United States, exclusively — production, backups and disaster recovery. We do not use offshore processing for any system that touches PHI, and our subprocessors are listed in writing under their own BAAs.
Do you train models on our data?
No. It is a contractual commitment, and it flows down to every model provider and subprocessor we use. Your data improves your configuration; it does not improve a shared model.
What happens if an agent makes a mistake?
It is logged, traceable and reversible. Below-threshold work never gets actioned in the first place, and every action carries the inputs that produced it — so you can see what happened, correct it and prove the correction.
Can we audit you?
Yes. We support your audit, your payer's audit, and your buyer's diligence. You can export the complete action log at any time, and we retain audit workpapers for the period your record-retention policy requires.
What is your breach notification commitment?
Notification without unreasonable delay and inside the regulatory window, with the scope, the affected records and the remediation already determined. We do not wait for a legal review to tell you something happened.
Do agents create a HIPAA risk we should document?
Yes, and it should be in your risk analysis. We provide the agent inventory, data flows and safeguards documentation you need for that analysis, and we will review it with your compliance lead.
Is there a vendor lock-in risk with AI?
Deliberately low. Configuration, prompts, routing rules and logs are yours and exportable in a standard format. The agents integrate through your EMR, so the systems of record stay yours either way.

Put us through your security review.

Send your security questionnaire and we will complete it before you sign anything — including the parts that are inconvenient to answer.

A 30-minute call with an operations lead. No obligation, no sales sequence, and we will say so if you do not need us.