Trust Center

HIPAA compliance and AI security.
Every claim verifiable.

Compliance isn't a checkbox. It's the product.

HASP is built on real governance infrastructure — not a BAA stapled to a chat wrapper. Every prompt is scanned for PHI before leaving your environment. Every action is hash-chained and Ed25519-signed. Every claim on this page has a verifiable artifact behind it.

Compliance · frameworks we ship under
Independently validated HIPAA BAA on every paid plan
Report under NDA SOC 2 Type II · attested
Certification in progress HITRUST CSF e1
EU + UK GDPR Art. 17 · 20 · 30
Active CCPA / CPRA California
Active + CPPA-ready PIPEDA Canada
Every AI request, traced

Eight enforcement steps between
your data and the model.

Every chat turn, uploaded text, and API call traverses the same pipeline. No surface can bypass it — not the UI, not the API, not the embedded Studio. PHI never reaches the model unprotected. Every step fires an audit event.

How it works

Four layers of HIPAA enforcement,
all the way down.

Each layer operates independently and is independently auditable. A failure at one layer doesn't silently collapse the others.

01

BAA enforcement at the gateway

Before any AI call is processed, the gateway checks that your organization has a countersigned, in-date Business Associate Agreement. No BAA — no AI traffic, full stop. The check runs at the gateway, not in application code, so it cannot be bypassed by any surface (UI, API, or embedded Studio).

View the BAA →
02

PHI scanning before every model call

Every chat turn, uploaded text, and tool definition passes through HASP's own PHI detection pipeline — with healthcare-specific custom recognizers — before any prompt leaves HASP. Your policy decides what happens next: send PHI to the model under your BAA, redact it before inference (and restore it in the response), or block the request entirely. Detection events are logged with the categories that fired, the action taken, and the user who initiated the request.

Trace the data flow →
03

Signed, hash-chained audit log

Every product-level event — chat turn, document upload, RAG retrieval, BAA accepted, API call, admin action — is an immutable entry in a hash-chained log. Each entry is signed with an Ed25519 key bound to your organization; the public key is published. Periodic checkpoints are anchored to an RFC 3161 Time Stamping Authority so timestamps are attested by an independent third party. The chain can be verified on your auditor's machine without any HASP software.

See the verification recipe → Walk through a compliance export →
04

Per-org dedicated data plane

Each Enterprise-tier organization gets a dedicated data plane — its own database, its own vector index, its own file storage, on dedicated infrastructure. There's no logical multi-tenancy at the data layer for Enterprise customers. Your data doesn't share compute with any other customer. Provisioning happens as part of Enterprise onboarding.

Data isolation details →
Compliance frameworks

One platform. Six frameworks covered.

HIPAA, GDPR / CCPA, and PIPEDA are platform-level commitments from V1 launch. HASP holds a SOC 2 Type II attestation — report available under mutual NDA — and HITRUST CSF e1 certification is in progress.

Platform-native

HIPAA

Business Associate Agreements available before signature. Control matrix, BAA, and data-flow inventory available to active prospects. The gateway enforces BAA status; the audit chain records every PHI-adjacent event. Customers are notified of a confirmed breach within 72 hours of discovery — well inside HIPAA's 60-day outer bound (45 CFR §164.404 / §164.410).

  • BAA available pre-signature
  • HIPAA Safe Harbor PHI categories enforced at gateway
  • Signed audit chain — every PHI-adjacent event logged
  • 72-hour customer breach notification (HIPAA outer bound: 60 days)
  • Safeguards independently validated through continuous monitoring
View BAA →
Platform-native

GDPR · CCPA / CPRA

Platform-level implementation from day one. Article 17 erasure (30-day offboarding window, full cascade), Article 20 portability (machine-readable export), Article 30 records of processing, ADM role boundaries. CCPA deletion and export served by the same mechanisms.

  • Right to erasure — 30-day cascade after offboarding window
  • Right to portability — machine-readable export available
  • Data Processing Agreement (DPA) available to all customers
  • 30-day sub-processor change notice
  • CCPA deletion + export via same mechanisms
Request DPA →
Platform-native

PIPEDA

Canada's federal private-sector privacy law. All ten PIPEDA fair-information principles are satisfied by the same controls that underpin GDPR — consent, access, erasure, portability, and sub-processor accountability. Cross-border transfer accountability under Schedule 1 §4.1.3 is discharged through contractual clauses in the DPA. The control set is also built to satisfy Canada's forthcoming CPPA (Bill C-27) when it comes into force.

  • Ten fair-information principles — same controls as GDPR
  • Cross-border transfer accountability via DPA §9 clauses
  • Individual access + challenge-compliance paths
  • Built to satisfy the forthcoming CPPA (Bill C-27)
PIPEDA details →
Attested · report under NDA

SOC 2 Type II

HASP holds a SOC 2 Type II attestation, with controls continuously validated through third-party compliance monitoring. The report and current bridge letter are available under mutual NDA. The compliance substrate additionally carries its own SOC 2 Type II covering the infrastructure layer HASP inherits.

  • HASP SOC 2 Type II — report + bridge letter under mutual NDA
  • Controls continuously validated by a third-party platform
  • Substrate SOC 2 Type II inherited at the infrastructure layer
SOC 2 details →
e1 certification in progress

HITRUST CSF

HASP's HITRUST CSF e1 certification is in progress. Today, HITRUST posture is inherited from our compliance substrate; the attestation letter is available under NDA on request.

  • HITRUST CSF e1 certification in progress
  • Substrate HITRUST posture inherited
  • Attestation letter available under NDA
HITRUST details →
Security posture

Infrastructure, auth, application,
and operational controls.

What follows answers the most common SIG / CAIQ / HECVAT questionnaire categories. Detailed questionnaire responses are available on request.

Infrastructure

Hosting
HIPAA-eligible compliance substrate on dedicated infrastructure. Global edge CDN, WAF, and DDoS mitigation at the edge.
Encryption in transit
TLS 1.2+ enforced at all boundaries (TLS 1.3 preferred). HTTPS-only; no plaintext fallback. The global edge network handles termination; internal traffic is re-encrypted.
Encryption at rest
AES-256 at the managed database layer. The compliance substrate's dedicated stack includes full-disk encryption on all compute and storage. OAuth tokens encrypted at the application layer.
Cryptographic modules
Security-load-bearing cryptographic operations performed inside the HASP platform route through FIPS 140-3 validated OpenSSL (the OpenSSL FIPS Provider 3.1.2, holding NIST CMVP certificate #4985, Level 1, Active through 2030). Covers outbound TLS (managed database, third-party APIs, internal traffic to the TLS-terminating load balancer), application-layer encryption (AES-256-GCM for HASP-managed keys and PHI maps; AES-256-CBC for framework-encrypted fields), password hashing (PBKDF2-HMAC-SHA-256), audit-chain hashing and security-relevant token hashes (SHA-256), and audit export signing (Ed25519). Operational fingerprints (cache keys, idempotency keys) and HMAC signature verification use FIPS-approved algorithms but are explicitly out of scope of the validated module. Inbound user-facing TLS is terminated at the substrate's TLS-terminating load balancer by a separately-validated cryptographic module — handled by the substrate, not by ours. Useful for DEA EPCS customers, where 21 CFR 1311.115 requires a validated cryptographic module. We do not claim FedRAMP or environment-level FIPS validation. See the full FIPS module verification recipe.
Backups
Substrate-managed automated backups with point-in-time recovery on all Enterprise-tier databases. Backup policy is per-org and independent of other tenants.
Availability
99.9% uptime target. 24/7 automated monitoring. Dedicated database cluster on Enterprise contracts. Status updates at status.usehasp.com.
IDS / vulnerability management
The compliance substrate includes intrusion detection and host-level vulnerability scanning. Findings feed into the substrate's operational SLA.

Authentication & access control

Authentication methods
Passwordless magic links, Google OAuth, Microsoft OAuth. Passwords are intentionally not supported — no credential stuffing surface.
Enterprise SSO
SAML 2.0 enterprise SSO. Okta, Azure AD, Google Workspace, and any SAML 2.0-compliant IdP supported. SSO enforcement is configurable per org; when enabled, OAuth and magic-link login are disabled for members.
SCIM provisioning
Automated user lifecycle management via SCIM 2.0. Available on Business and Enterprise plans.
Role-based access
Three org roles (Owner, Admin, Member) plus per-app roles (Admin, Editor, Viewer). Access to internal apps is scoped at publish time; end-users only see what they're granted.
Session security
CSRF protection on all state-changing requests. Secure, HttpOnly session cookies. Automatic session expiry. Sessions are audited on impersonation.
Internal access
Founding team only; no broad support access. Impersonation is audited and visible in the customer's own audit log. Production database access is through the compliance substrate's audited access controls.

Application security

XSS prevention
Strict Content Security Policy with per-request nonces. User-uploaded apps execute on an isolated domain (usehasp.run) so any breakout is contained to that sandbox.
CSRF protection
Automatic CSRF token verification on all state-changing requests. Framework-enforced; cannot be bypassed per-route.
Input validation
Server-side validation on all inputs via dedicated validation classes. Client-side validation is layered on top, not relied on alone.
Rate limiting
Per-user and per-org rate limits on all API endpoints. Tiered by plan. Gateway-level throttling at the global edge network before requests reach the application.
File upload security
File type validation, size limits, and MIME sniffing prevention. Document uploads are processed server-side and never served back from the origin.
Security headers
HSTS, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, and Permissions-Policy applied globally. CSP nonce per request.

Operational security

Secrets management
Centralized secrets vault for all environments. No secrets in source code or environment files. Rotation is tracked. Production secrets are not accessible to development environments.
CI / CD
Automated CI/CD pipeline with path-filtered workflows. Lint, migration safety checks, and test suites run on every PR. Deployments are gated behind branch protection and require passing CI.
Dependency management
Automated dependency scanning with daily updates. Vulnerability audits run on every build. Critical CVEs are addressed promptly, with priority by severity.
Incident response
Affected customers are notified of a confirmed breach within 72 hours of discovery — well inside HIPAA's 60-day outer bound (45 CFR §164.404 / §164.410). GDPR Article 33 (72 hours to supervisory authority). Customer-impacting incidents reported on status page and via email promptly upon confirmation.
Data lifecycle
30-day offboarding window before any org-wide deletion cascade. Everything exportable in machine-readable form during that window. After 30 days, PHI is deleted from primary storage; backup exclusion enforced on restored data.
Vendor management
All sub-processors documented in the register below and in the DPA. Sub-processor changes carry a 30-day advance-notice obligation. All key vendors hold SOC 2 certifications.
Sub-processor register

Every downstream processor. No surprises.

Categories of downstream processors. The complete vendor-named register lives in the sub-processor register and the DPA. Changes carry a 30-day advance-notice obligation under the DPA. You can object to a new sub-processor before any change takes effect. Last updated: May 2026.

PHI handling is HASP-owned. PHI de-identification, redaction, and re-identification is performed by HASP's own anonymization pipeline with healthcare-specific custom recognizers — not delegated to any third party. Inference flows through direct provider integration under HASP-direct BAAs; when your org's policy elects redaction, PHI scanning and de-identification run before content leaves HASP's substrate.
Capability Role Purpose Data in scope Location
HIPAA-eligible compliance substrate Infra Managed database, application hosting (compute + network); HIPAA + SOC 2 substrate inheritance Application data, per-org databases, audit infrastructure United States (dedicated infrastructure)
Inference providers AI Inference for chat, document analysis, AI Studio, public API, and agent workflows — direct integrations under HASP-direct BAAs. See the sub-processor list for named providers and the model catalog for supported models. Prompt + completion content, per your org's PHI policy — sent under a HASP-direct BAA, or automatically de-identified when your policy elects redaction United States
Embedding provider AI Document embeddings for RAG Document text after PHI redaction United States
Web search provider (primary) AI Web search retrieval for the AI's web.search tool Search queries only. HASP ensures no PHI is ever transmitted to this provider. United States
Web search provider (fallback) AI Web search retrieval — optional fallback Same scope as primary. HASP ensures no PHI is ever transmitted to this provider. United States
Edge CDN / object storage Infra CDN, object storage, DNS, SSL, custom domain routing (usehasp.run) Request metadata only — IP, TLS handshake, User-Agent. No PHI at edge. Global (US-headquartered)
Payment processor Business Subscription billing, usage-based billing reporting, payment processing, tax calculation Customer billing metadata; no PHI United States
Transactional email provider Business Transactional email delivery Email addresses, transactional message content; no PHI United States
Secrets management Business Secrets management — runtime configuration, encryption keys, third-party API credentials Application secrets and configuration. No customer data, no PHI. United States
Enterprise SSO (SAML) Business SAML 2.0 SSO, SCIM provisioning (Business + Enterprise tiers) Identity attributes, group memberships. No PHI. United States
OAuth identity providers Business OAuth authentication (when Customer enables provider sign-in) Authentication identity attributes; no PHI United States
Error tracking Analytics Application error monitoring and performance telemetry Application telemetry; PHI scrubbed at source United States
Application performance monitoring Analytics Application performance monitoring — job queue health, slow queries, scheduled commands, server metrics Application telemetry; PHI scrubbed at source United States
Product analytics Analytics Product analytics (only when end user consents via cookie banner) Marketing-site and consented in-app behavior. No PHI. United States
The standard we hold ourselves to

What we show that our competitors can't.

Most HIPAA-compliant AI vendors are a BAA and a chat UI. We built governance infrastructure that holds up the first time you get audited — and we can prove it. Here's what distinguishes real compliance from a checkbox.

Auditor-verifiable chain

The audit log can be verified on your auditor's own machine with no HASP software in the loop. Download the export, run the script, confirm both the hash chain and each Ed25519 signature. One command does it — our open-source CLI: npx @usehasp/verify export.json. The verification recipe is public. Few competing vendors offer independently verifiable audit exports — most audit trails are screenshots and CSVs you take on faith.

Verification recipe →

PHI gateway with a traceable audit event

PHI scanning fires an audit event with the entity categories that triggered, the action taken (de-identify / allow / block), and the user who initiated it. You can answer "was any PHI in that conversation?" in 30 seconds from an audit export. We haven't found another HIPAA AI vendor that provides this at the application-log level.

Data-flow inventory →

Internal-app builder on the same BAA

HASP pairs HIPAA-compliant AI with a full internal-app builder — a combination we haven't seen anywhere else. The tools your team publishes — intake forms, care-coordination dashboards, referral workflows — run under the same BAA, on the same audit trail, on the same bill. Typical HIPAA AI wrappers have no tool-building capability; the general-purpose internal-tool builders aren't HIPAA-compliant.

Per-org dedicated data plane

Every Enterprise-tier org runs on its own dedicated data plane on dedicated infrastructure — a separate database cluster, its own vector index, and its own object-storage bucket. Cluster-level isolation your security team can point to on an architecture diagram.

Data-flow inventory mapped to questionnaire answers

The data-flow inventory traces every hop from browser to storage, with per-hop controls, retention curves, and BAA scope. It's written to map directly to SIG, CAIQ, and HECVAT question categories — so your procurement team isn't building this from scratch.

Read the data-flow inventory →

Audit chain integrity is not a paid upgrade

Most compliance AI vendors gate their strongest audit features — tamper-evidence, signed exports, extended retention — behind enterprise tiers. HASP does not. The signed audit chain, Ed25519 verification, and RFC 3161 timestamp anchoring apply to every paid plan, including Solo. A solo practitioner gets the same audit guarantees as a tier-1 health system. Higher tiers add analytics depth and retention windows — not the integrity of the chain itself.

Multi-provider BAA through one vendor

HASP holds direct BAAs with every inference provider in the path — and exposes them through one HASP BAA you sign. As we add inference providers, your BAA doesn't change; we handle the chain downstream and notify you 30 days before any sub-processor change takes effect. No third-party AI gateway in the path.

For procurement teams

Need the full package?

Security questionnaires (SIG, CAIQ, HECVAT, or customer-specific forms) are completed promptly. The SOC 2 substrate report, penetration-test summary, BAA, DPA, attestation letters, and vendor-specific security overview PDFs are available under a mutual NDA.

Other contact paths →
NDA available on request.