You're shipping GenAI applications faster than you can audit them. Your agents are multiplying across cloud providers without central oversight. Your teams are juggling manual policy checks while compliance gaps slip through the cracks.
Reactive governance doesn't scale. When your AI portfolio spans hundreds of models and agents across multiple clouds, manual compliance checking becomes a bottleneck that slows innovation while creating blind spots. Teams need systematic policy enforcement that keeps pace with deployment velocity.
This release transforms Arthur into a comprehensive governance platform. From automated policy compliance chains to agent discovery across cloud providers, from bulk workspace auditing to persistent audit logging, May 2026 delivers the infrastructure your organization needs to govern AI at scale without slowing down development.
Policy-Driven Compliance Automation
Manual compliance checking breaks down when you're managing policies across dozens of models and agents. Teams waste hours on repetitive rule validation. Violations slip through when policy checks lag behind deployment cycles.
Arthur now automates compliance from end to end. Bulk workspace compliance checking triggers evaluation across every policy-model pair in a workspace with a single API call, executing a complete Metrics → Alerts → Compliance chain to ensure fresh data at every stage. NEEDS_ATTENTION status provides a remediation grace period before escalating to NON_COMPLIANT, while webhook notifications fire automatically when violations exceed enforcement delays.
PMs get governance that scales: Policy assignments materialize alert rules automatically, compliance history provides infinite-scroll audit trails, and workspace compliance overview delivers a flat, queryable table showing per-rule status across all policies.
Developers get flexibility: Configurable alert intervals prevent false positives, validation endpoints test queries before deployment, and the compliance API supports both bulk workspace checks and granular model-specific evaluation.
Compliance teams get visibility: Three new governance roles enable delegation without bottlenecks, audit logging captures every API request with full user context, and persistent audit storage ensures durable compliance records.
Database Performance at Scale
Database queries that worked fine with hundreds of models start timing out at thousands. Job polling becomes a bottleneck. API latency degrades user experience.
This release delivers fundamental performance improvements: optimized authentication queries eliminate unnecessary user table writes on every request, while partial dequeue indexing reduces job polling queries from 225ms to 0.008ms with minimal buffer access.
These changes ensure Arthur scales smoothly from hundreds to thousands of monitored models without degrading performance.
Enterprise Infrastructure Reliability
Production deployments need persistent audit storage, high-availability databases, and support for security-hardened environments like OpenShift.
Persistent audit log storage for Kubernetes deployments ensures compliance records survive pod restarts and cluster migrations. Standardized Helm charts provide consistent deployment patterns across control plane and stateful services, while OpenShift support includes proper fsGroup configuration and audit PVC management.
Azure Blob Storage connector expands data source compatibility, enabling teams to monitor models regardless of their storage infrastructure.
Alert Lifecycle Observability
When alerts fire and resolve without clear tracking, teams lose visibility into system health patterns. Alert debugging becomes guesswork.
Alert lifecycle logging now tracks when alerts fire, resolve, and fail to compute due to missing data. Enhanced documentation covers the complete alert system architecture, enabling teams to monitor alert resolution times and diagnose data availability issues proactively.
From Reactive Compliance to Proactive Governance
From scattered agents to centralized discovery. From manual policy checks to automated compliance chains. From performance bottlenecks to optimized queries at scale.
Arthur transforms governance from a development blocker into an enabler. Teams ship faster knowing their AI applications are continuously monitored, compliant, and discoverable across any cloud infrastructure.
PS — The database performance improvements alone make this release worth upgrading for. Seeing job polling drop from 225ms to sub-10ms is the kind of change that makes the platform feel completely different at scale. Having struggled with slow API responses on large Arthur deployments myself, these optimizations should eliminate a major pain point for anyone running hundreds of models. Drop us a line at team@arthur.ai if you're seeing significant performance improvements after upgrading — We’d love to hear about it. See the full platform release notes for May 2026 here: https://docs.arthur.ai/changelog
SHARE
