Skip to main content
Trust Center

Security posture, in plain detail.

L2H is engineered for regulated, enterprise, and government environments. Architecture and controls map to NIST 800-53 Rev. 5, NIST AI RMF, NIST CSF, SOC 2 Trust Service Criteria, and ISO 27001 control families. Append-only audit log, scoped + expiring API keys, signed webhooks, workload-identity-bound cloud access, secrets-manager integration, and hardened containers — not a hand-wavy “we have RBAC.”

Compliance posture

Architected to the frameworks that matter.

Controls and architecture mapped to enterprise and federal frameworks. Two tracks plus a shared core — pick the one that matches your buying motion.

Commercial & Enterprise

AWS (Commercial)

In production

Customer-hosted AWS deployments — your account, your VPC, your data. Workload-identity-bound cloud access and secret-store integration.

Azure (Commercial)

In production

Customer-hosted Azure deployments for Microsoft-aligned customers. Azure AD / Entra ID SSO supported.

SOC 2 Type II

Roadmap

Architecture and controls designed against SOC 2 Trust Service Criteria. A Type II audit is on the roadmap; we are not currently SOC 2 attested.

HIPAA

Architecture support

L2H provides administrative, technical, and physical safeguards designed to support a covered entity's or business associate's HIPAA obligations. L2H itself does not act as a HIPAA-covered entity.

PCI DSS

Architecture support

L2H is not in scope as a card-data processor and does not hold a PCI DSS attestation. Customers handling cardholder data are responsible for their own assessment.

ISO 27001 / 27701 / 42001

Roadmap

Architecture and ISMS controls mapped to ISO 27001, 27701, and 42001 control families. L2H is not currently ISO certified.

GDPR

Architecture support

Data Processing Addendum, Records of Processing Activities, and Standard Contractual Clauses available on request to support customer GDPR obligations. Customers act as the data controller for content processed in their environment.

Federal & Defense

FedRAMP High (control baseline)

Architecture support

Architecture and controls designed against the FedRAMP High control baseline (NIST 800-53 Rev. 5 High). L2H does not currently hold a FedRAMP authorization. Use of the FedRAMP name does not imply FedRAMP authorization.

NIST 800-53 (High)

Architecture support

Mapped to NIST 800-53 Rev. 5 High baseline controls.

DoD CC SRG IL4

Architecture support

Designed to support customer deployment in IL4-eligible Azure GovCloud (DoD/DISA) environments. Authorization status is held by the deploying customer / sponsoring agency.

DoD CC SRG IL5

Architecture support

Designed to support customer deployment in IL5-eligible environments using Azure GovCloud (DoD/DISA) and private routing. Authorization status is held by the deploying customer / sponsoring agency.

JWICS / TS/SCI patterns

Architecture support

Designed to support deployment in air-gapped TS/SCI environments using on-prem Kubernetes and self-hosted LLMs. Accreditation is held by the sponsoring agency.

DoD CC SRG IL6+

Architecture support

Air-gapped on-prem deployment pattern with self-hosted LLMs and self-hosted IDP, designed to map to IL6+ control expectations.

Air-Gapped Deployment

In production

On-prem Kubernetes, self-hosted LLMs, zero external dependencies at runtime.

Shared across both tracks

NIST AI RMF

Architecture support

Risk-management practices mapped against the NIST AI Risk Management Framework.

NIST CSF

Architecture support

Cybersecurity Framework controls implemented across identify / protect / detect / respond / recover functions.

OWASP ASVS

Architecture support

Application Security Verification Standard hardening practices applied to the platform.

Status reflects how the L2H architecture is designed against the named control frameworks. L2H does not currently hold SOC 2, ISO, or FedRAMP attestations and is not on the FedRAMP Marketplace. Authorization status of any deployment in DoD or IC environments is held by the deploying customer or sponsoring agency. Trademarks referenced (FedRAMP, NIST, ISO, AWS, Azure, ServiceNow, others) are property of their respective owners; reference does not imply endorsement, sponsorship, or affiliation.

Security architecture

Ten layers, all shipping today.

Append-only audit log

Comprehensive event catalog covering templates, runs, schedules, integrations, LLM connections, plugins, webhooks, teams, API keys, and auth providers. UI with category pills, full-text search, pagination, CSV export. Configurable retention.

API keys with scope + expiry

High-entropy keys hashed at rest. Per-key scopes, expirations, and last-used metadata. Immutable after creation; rotation via deactivate-then-create.

Signed outbound webhooks

Industry-standard HMAC signing with replay-resistant timestamps. SSRF guard blocks loopback, private ranges, link-local, IPv6 ULA, and cloud-metadata addresses. Bounded retry policy. Secret returned only once at creation.

Rate limiting + idempotency

Per-key and per-user token-bucket limits across run dispatch and streaming. Idempotency keys supported on run POSTs.

Site-admin RBAC

Site-admin status synced from your identity provider on every login. Site admins manage plugins, auth providers, global workflows, and embedded-assistant config.

Embedded-assistant contract

Tightly scoped, in-process only. The embedded assistant cannot exceed the triggering user's permissions and is never accepted from outside the cluster.

Defense in depth

Hardened, non-root containers, capability stripping, network-policy templates, and read-only filesystems where compatible.

Secrets management

Secrets pulled from your cloud secret store at process start. Never embedded in container images. LLM and integration credentials are referenced indirectly so catalogs never store raw keys.

Web hardening

Strict CSP, hardened response headers (X-Content-Type-Options, X-Frame-Options, Referrer-Policy, Permissions-Policy), optional HSTS, and server-side auth tokens.

Identity contracts

Standards-based authentication for both management and run-time surfaces, with separate guards for each tier. Embedded-assistant access is in-process only.

Production safety guards

Hard refusals, by design.

  • The API refuses to start with auth-bypass flags enabled in production.
  • Deployment notes warn on production misconfiguration.
  • Audit logging is fail-soft — failures cannot roll back real mutations.
  • Outbound HTTP paths sanitize header values to prevent header injection.

Identity providers supported

SSO, M2M, mTLS.

  • OIDC (Auth0, Azure AD, Okta, Keycloak, Google)
  • SAML 2.0
  • LDAP / Active Directory
  • Local accounts
  • PKI / client certificate (mTLS)
  • Hashed, scoped API keys

One identity across Orchestrator + Chat. Same Auth0/IDP tenant → SSO “just works.”

Downloadable artifacts

For your security review.

Security Whitepaper

Architectural security overview, threat model, control families addressed.

Request copy

NIST AI RMF Mapping

How L2H controls map to the NIST AI Risk Management Framework.

Request copy

NIST 800-53 Control Mapping

Control-by-control mapping to the NIST 800-53 Rev. 5 High baseline.

Request (NDA)

Data Processing Addendum

Standard L2H DPA template, available on request.

Request copy

Subprocessors List

Current subprocessors with subscribe-to-updates.

Request copy

Vulnerability Disclosure Policy

How to responsibly report security issues.

Request copy

Have a security question?

Send us your security questionnaire, ATO support request, or vulnerability disclosure. We respond within one business day.