Security & Compliance

Security built to be reviewed.

A dedicated environment for your organization. Multi-factor authentication for every user in production. PHI behind its own permission and hidden on screen by default. Every access recorded in your own audit trail. Each control is described below in the terms your security team will ask about.

dedicated environment per organization MFA required in production 15roles 162audited actions U.S.-hosted
15
Granular roles
7 carry PHI access, 8 do not — export is a permission again
162
Audited actions
Written before the response returns
30 min
Idle timeout
Automatic sign-out of an inactive session
3 hr
PHI reveal expiry
A reveal lapses on its own, and is logged
7 yr
Audit retention
2,555-day default, above HIPAA's six-year minimum
Data Isolation

Your Organization's Own Environment

Your members are not rows in a table shared with other customers. Each organization gets its own environment when we provision it.

A dedicated analytics environment

Your claims, eligibility, quality and member data are held in an analytics dataset created for your organization alone, with its own audit table. There is no shared member table and no query path from one organization's environment into another's.

Dedicated file storage

Documents, exports, inbound data files and call recordings are held in storage scoped to your organization. A file-transfer account is confined to your organization's own area, and attempts to traverse out of it are refused.

Your own identity configuration

Sign-in is configured per organization on an enterprise managed identity service. Your users exist in your organization's own identity configuration, and a session carries the organization it belongs to as a signed claim.

Every request bound to your organization

Before a request reaches application code, the platform verifies the session signature and checks that the organization in the session matches the address being used. A mismatch is refused before any data is read, and a request that cannot be resolved to an organization fails closed.

Separation from the first day

Isolation is established when your environment is provisioned — dataset, storage, identity configuration and audit table — rather than configured after your data arrives.

An organization can hold many clients

A TPA, coalition or advisory firm can hold many client environments under one organization, with its own administrators and cross-client work queues. Each client's data stays inside its own environment.

Identity & Access

Identity, Roles, Least Privilege

Who someone is, what their role permits, and whether that role may see a member identity are three separate questions. The platform answers them separately.

Multi-factor authentication in production

MFA is required for every user on every organization in production — including every privileged role. The second factor is a code sent to the user's phone.

Step-up re-authentication on sensitive actions

Running a data export, changing an existing user's role, or changing trusted-device policy requires an authentication completed within the last five minutes. The check is enforced on the server against a signed, user-bound cookie, not in the browser.

15 roles, with PHI as its own permission

Seven of the fifteen roles carry PHI access and eight do not; the ability to export is a separate permission again. A finance lead can work the cost data without member identities, and an executive can review program performance with identities hidden.

PHI hidden by default, every user, every session

Names, dates of birth and identifiers render hidden on every screen. Only a role with PHI access sees a reveal control; revealing is an explicit action, never a side effect; the reveal expires after three hours; and each one is written to the audit trail. Most analytical work never needs the data unmasked.

Your administrators can tighten any role

PHI access, export permission, dashboards and features can be further restricted per role for your organization, from your own administration console. Platform super-admin permissions cannot be widened or weakened by an organization-level override.

Automatic logoff

An inactive session is signed out after 30 minutes, which is the control HIPAA's automatic logoff specification asks for. The timeout is configurable for your organization.

Trusted devices, centrally revocable

A device that has passed MFA can skip the prompt for a period you set, capped at 30 days. Trust for a user is revoked centrally, which invalidates every device already trusted for that user.

Account setup and recovery are rate-limited

Account-setup links are single-use and expiring. Setup, password-reset, trusted-device and step-up endpoints are rate-limited centrally rather than per server, so the limit holds however the traffic is distributed.

No PHI without an executed BAA

PHI-serving routes check that an executed Business Associate Agreement is on file for your organization and refuse the request if it is not. The agreement is a gate in the code, not only a line in a contract.

Named identities for machines too

Scheduled jobs and internal services authenticate as named service identities with signed tokens, scoped to the one service they are allowed to call. Inbound webhooks from our telephony and alerting integrations are signature- or credential-verified, and rejections are logged.

Encryption

Encryption in Transit and at Rest

Encryption is the easiest claim to make and the easiest to overstate, so here it is precisely.

In transit

Browser sessions and API traffic reach the platform over HTTPS at a managed load balancer. Data feeds arrive over an encrypted file-transfer channel. Calls from the platform out to our cloud, email, telephony and AI services are made over HTTPS.

At rest

Everything stored — the analytics environment, documents, inbound files, recordings and logs — is encrypted at rest with AES-256 by the cloud platform's default encryption, under keys the provider manages and rotates. The application never handles raw key material.

Credentials and keys

Credentials, API keys and integration secrets are held in a managed secret store and injected at deploy time. They are not in source code, not in configuration files, and not in the container image.

Controlled egress

All outbound traffic from the platform leaves through a single controlled egress path with a fixed address, so your network team can allowlist it and see exactly where platform traffic originates.

PHI responses are never cached

Responses that carry member data are marked private and no-store, so they are not held in a browser cache or an intermediate cache.

Exports are deliberate

Export is a role permission, an export job requires fresh re-authentication, and document downloads are issued as links that expire in 15 minutes. Spreadsheet exports are written so that a cell beginning with a formula character cannot execute when opened.

Audit Logging

Every Access, Recorded

The audit trail is where a compliance reviewer spends the most time, so this is what it does and where it lives.

162 audited action types

Authentication, MFA, step-up and trusted-device events; PHI access and exports; administrative and role changes; outreach activity; care-gap lifecycle; and every AI query. The list is a fixed taxonomy in the code, not free text a caller chooses.

Written before the answer is returned

On PHI and administrative routes, audit rows are flushed before the response leaves the server rather than queued for afterwards — so the record exists even if the session ends the moment the data appears.

Held in your own environment

Audit rows are written to an audit table inside your organization's own analytics environment, and in parallel to the platform's operational log for real-time monitoring and alerting.

What a row contains

The user and their role, the organization, the action, the resource, the timestamp, the source address, the user agent and the session. Event details pass through a redactor before they are stored, and member identifiers are masked in the detail field.

Seven-year default retention

The retention setting for your organization's audit table defaults to 2,555 days — seven years, above HIPAA's six-year documentation requirement. You can set a different period, with a 30-day floor the platform enforces.

Available to your compliance team

The trail is queryable for your own compliance reviews, HIPAA audit support and access reviews. We will walk your privacy officer through it during onboarding.

AI Safeguards

AI Under the Same Controls

AI is a standard part of a security review now, and rightly so. The platform's AI features are consumers of the controls above, not exceptions to them.

Grounded in your environment

An AI answer is computed from your organization's own data. Another customer's data is not part of the request.

The asking user's permissions apply

The role and PHI permissions of the person asking govern what the model can be handed, so the AI cannot surface something that user could not open directly. Masking still applies to what comes back on screen.

Every AI data access is audited

Each AI tool call and chat request writes a row to the audit log of the organization whose data was read, naming the user, the surface, the tool, the data class, the result size and the outcome. The row has no field that could carry prompt text, model output or raw results — and the browser cannot write these records at all, so the trail cannot be forged from a client.

Not used to train models

AI features run on our cloud provider's enterprise AI service, covered by the same signed Business Associate Agreement as the rest of our cloud infrastructure. Under that service's contractual terms, your data is not used to train or improve the provider's models.

Care gaps are specification logic, not inference

No model decides whether a care gap is open. All 71 quality measures follow their published specifications, return the same answer every time, and can be checked line by line against the spec. AI is used to prioritize, summarize and draft — never to decide what counts as a gap.

AI calling and AI texting are switched off

AI-generated voice calling and AI-generated text messaging are disabled across the platform — in the dispatcher, in the campaign routes and in the interface. Outreach is a coach's own call, an email, or the care reminder texts described below.

Member Messaging

Text Messages Carry No PHI

The care reminder program is deliberately built so that a text message on a lock screen discloses nothing.

No clinical detail in the messageMessages are neutral nudges. No diagnosis, medication, test result, provider name or appointment time appears in the body. The message directs the member to sign in, where anything sensitive sits behind authentication.
Consent is the plan's, documentedThe health plan obtains express consent at enrollment, and our agreements require documented consent before a member's information reaches us for outreach. No more than 2–4 messages a month, STOP at any time.
Mobile data is never soldPhone numbers and SMS opt-in data are never shared or sold to third parties or affiliates for marketing. They go only to the subprocessor that transmits the message, under written terms.

The full program terms, including the exact opt-in, STOP, HELP and frequency language, are in our Privacy Policy, section 3.

Infrastructure

Infrastructure and Resilience

We do not run our own data centers. The platform is built on healthcare-eligible cloud services, under an agreement that puts HIPAA obligations on the provider.

HIPAA-eligible cloud services under a signed BAA

The platform runs on Google Cloud, using services designated eligible for healthcare workloads under the Business Associate Agreement we have signed with Google. That agreement covers the services that store and process your data, and the identity and AI services the platform depends on.

United States only

Your data is stored and processed in United States cloud regions. We do not operate the platform from outside the United States, and we do not move your PHI offshore to deliver the service.

Private by default

Two surfaces face your organization: the web application and the file gateway for data delivery. Nothing else is exposed — the analytics environment, internal services and the secret store are reachable only by the platform's own authenticated service identities.

Managed, auto-scaling capacity

The application runs on managed infrastructure that scales with demand, and stores data on managed cloud services. There is no server we patch by hand and no maintenance window for capacity.

Monitored, with alerts to the engineers

Platform health, data pipeline failures and security-relevant integration failures raise alerts to our engineering team, and request identifiers tie a user report to the exact request in the logs.

Validated at intake, nothing loaded blind

Format, structure and content checks run automatically when a file arrives. Records that fail are held in a quarantine queue with per-row error reporting for your data team to review — invalid records are neither silently dropped nor silently loaded. A measure whose source data is absent is marked as such and never run, so an empty feed cannot manufacture phantom gaps.

Secure Development

How Changes Reach Production

A control is only as good as the change process around it, so this is ours.

Reviewed before it merges

Every change goes through pull request review and is traceable to an author and an approver. An automated review gate runs over the staged change before a commit is accepted, in addition to the human review.

One automated path to production

Builds and deployments run through an automated pipeline with its own record. A release is first deployed as a revision that receives no traffic; promoting it to production is a separate, deliberate step taken by a person. There is no route for a hand-built artifact or an ad-hoc production change.

Regression tests in the pipeline

An automated regression gate runs in continuous integration before a release can proceed, covering authentication, care gap, engagement, forms, financial and billing behavior along with the platform's type contracts.

Penetration tested, and honestly described

In April 2026 the production site and all of its subdomains were penetration tested under written authorization from the platform owner. The test was gray-box: unauthenticated external testing against the live environment alongside source-code-informed review of the security-critical paths. Each finding is written up with severity, reproduction steps, affected files and a recommended fix, and tracked to remediation. We will discuss the scope, the methodology and the current remediation status with your security team.

Controls re-checked every release

We re-run an automated HIPAA control audit over the codebase after each release and record its result. The technical safeguard documents are reviewed monthly against the code, and the full program — administrative safeguards included — quarterly.

Change records, not recollection

Security configuration changes inside the platform — roles, permissions, security settings, user lifecycle — are themselves audited actions, so a review question about who changed what has an answer in the trail rather than in someone's memory.

HIPAA Program

HIPAA Program and Agreements

We are your Business Associate. Several of the commitments a reviewer cares about most are contractual, and we say which agreement carries them.

A BAA before any data moves

We sign a Business Associate Agreement with every customer before PHI moves, and the platform enforces it: PHI-serving routes refuse an organization with no executed BAA on file.

A DPA alongside it

A Data Processing Agreement sets retention periods by data class, the deletion timetable on termination, and subprocessor terms in writing — so the lifecycle is governed by your agreement rather than by a platform default.

A documented program, not a slide

Our HIPAA program is a set of thirty maintained documents covering administrative and technical safeguards — access control, audit logging, authentication, encryption, incident response, breach notification, contingency planning, vendor management and the rest — each mapped to the controls in the code.

Incident response

A documented process covers detection, containment, preservation of evidence, customer communication and remediation. Audit records and request identifiers let us reconstruct what was accessed rather than estimate it.

Breach notification, per HIPAA and your BAA

Every suspected impermissible use or disclosure is assessed using the four-factor risk assessment the Breach Notification Rule requires, and the assessment is documented either way. HIPAA requires a business associate to notify the covered entity without unreasonable delay and no later than 60 days after discovery; where your BAA sets a shorter deadline, the BAA controls.

Your data stays yours

Your organization retains ownership of its data. We process it to provide the service, as your business associate. It is not sold, not shared between customers, and not used to train AI models.

Vendor Security Review

Ready for your vendor review?

Send your security questionnaire to contact@healthbeniq.com and we will answer it. For a health plan's formal vendor review we can also provide, under NDA:

  • Our HIPAA control matrix, mapped to the 45 CFR Part 164 safeguards
  • Our Business Associate Agreement and Data Processing Agreement
  • Our subprocessor list, by category and by vendor
  • A walkthrough of the controls on this page with the engineers who built them
  • Written answers to your own security questionnaire