ChangelogWhat's new in Obiguard

Changelog.
Every release.

A complete record of every version shipped — new detectors, platform features, certifications, and fixes.

v 2.4.1

patchJune 2026
  • Fixed a race condition in the tool-call validation path under high concurrency.
  • Improved p95 latency for Bedrock-routed requests by 3ms.
  • Corrected EU AI Act Art. 13 mapping for the toxicity detector family.

v 2.4

minorMay 2026
  • Shipped 8 new detectors: IBAN, NHS number, Australian TFN, and 5 semantic injection variants.
  • Added OpenTelemetry export for metrics and traces to Datadog, Grafana, and any OTEL backend.
  • Policy Sets now support multi-agent assignment — one set, many agents, one edit.
  • Review Queue redesign: bulk assignment, keyboard-driven review, and annotation export.
  • Framework mapping updated for ISO 42001:2023 revision.

v 2.3

minorMarch 2026
  • GA: on-premises deployment with VPC-native enclave attestation.
  • Streaming inspection support — violations fire mid-stream, not only on completion.
  • Added Splunk HEC and Sumo Logic native connectors for audit log export.
  • AI Use Case registry — link agents to business functions and owners.
  • Launched EU AI Act control mapping across 47 detectors.

v 2.2

minorJanuary 2026
  • SOC 2 Type II certification achieved for the SaaS gateway.
  • ISO 42001 certification achieved.
  • Added Google Vertex AI and Azure OpenAI provider support.
  • Framework mapping now covers NIST AI RMF, SOC 2 CC6/CC7, ISO 42001, and HIPAA.
  • API-first policy management: all resources manageable via REST API.

v 2.0

majorOctober 2025
  • Major platform release: Policy Sets, Criteria, and AI Agent registry.
  • Human-in-the-loop Review Queue — assign, annotate, escalate violations.
  • Immutable Audit Ledger with cryptographic signing and SIEM export.
  • Dashboard redesign with real-time violations feed and cost analytics.
  • 64 detectors at GA covering PII, injection, jailbreaks, and tool-call scope.