Changelog

The admin portal now includes a dedicated view for webhook delivery status. Every webhook event shows delivery attempts, response codes, retry history, and next scheduled retry. Failed deliveries surface immediately — no more guessing whether your SIEM received the alert.

The model health system now tracks provider-level error budgets alongside latency and availability. When a provider approaches its error budget threshold, the circuit breaker activates preemptively rather than waiting for consecutive failures. Dashboard surfaces budget consumption in real time.

New dedicated landing pages for Okta and Microsoft Entra ID integrations. Each page walks through the complete setup — SAML 2.0 SSO configuration, SCIM 2.0 provisioning, and group-to-role mapping — with compliance callouts for organizations operating under SOC 2 or HIPAA requirements.

New engineering blog post explaining why granular DLP detection matters. Covers the 39+ individually-typed credential patterns, per-type policy rules (BLOCK vs REDACT vs LOG), and how entity-level detection feeds compliance reporting for PCI-DSS and HIPAA.

Hybrid Outpost deployments now export additional Prometheus metrics covering DLP inference latency, policy evaluation time, and local audit log volume. These metrics integrate with existing Grafana dashboard definitions, giving operations teams visibility into data-plane performance without routing telemetry through the cloud control plane.

Users can now initiate their own account recovery through a verified backup email or recovery code flow. The self-service path reduces administrator workload while maintaining full audit trail coverage — every recovery attempt, successful or not, is logged with device fingerprint and timestamp.

New middleware layer enforces passkey-only authentication for designated user groups. Organizations can require WebAuthn/FIDO2 credentials for admin roles, compliance-sensitive teams, or entire tenants. Fallback to password is blocked at the middleware level — not just hidden in the UI.

The admin portal now supports server-side pagination across all list views — users, audit logs, DLP events, and policy rules. Pages load consistently regardless of dataset size. Cursor-based pagination replaces the previous offset approach for more stable results during active data changes.

Every data-fetching view in the admin portal now renders a skeleton loading state while waiting for the API response. Previously, some views showed a blank area during load. The skeleton states match the final layout, preventing content shift when data arrives.

All page titles and meta descriptions across the public site have been optimized for search engine display. Titles now consistently include primary keywords and the Arbitex brand name within the recommended character range. Descriptions provide clear value propositions within display limits.

New engineering blog post covering how to run a complete DLP pipeline, policy engine, and audit system inside air-gapped networks. Covers Outpost architecture for defense, classified research, healthcare, and financial services environments that require zero external connectivity.

Administrators can now initiate account recovery for users who lose access to their primary authentication method. Recovery flows generate time-limited, single-use tokens with full audit trail coverage — every recovery action is logged with the initiating admin, target user, and timestamp.

New admin controls for tenant provisioning, suspension, and decommissioning. Suspended tenants retain data but lose API access. Decommissioned tenants enter a 30-day retention window before permanent deletion. All lifecycle transitions are audit-logged.

Budget caps and usage quotas now support per-team and per-department scoping in addition to per-user limits. Quota thresholds trigger configurable alerts at 75%, 90%, and 100% consumption. Exceeded quotas block new requests with a clear error message — no silent failures.

Resolved three edge cases where Outpost deployments could report stale pattern versions after an update cycle. The sync verification now includes a checksum comparison between the control plane pattern manifest and the local Outpost ruleset, ensuring exact parity.