Releases, milestones and updates.

Xybern has released the Agent Authorization Benchmark (XAAB), an open, reproducible benchmark of 137 attack and legitimate scenarios. The Xybern Authorisation Layer caught 100% of unsafe agent actions with zero false positives, and held at 100% under rephrasing where a strong keyword guardrail collapsed from 84% to 44%.

Xybern Authorised Agents is now fully available in Modern Standard Arabic, enabling enterprises in Saudi Arabia and the UAE to build, run, and govern AI agents entirely in Arabic. The release covers Skills, Automations, and Connectors with the full authorisation layer localised for Gulf markets.

Xybern Authorization Layer, the AI agent enforcement and governance control plane, is now fully localized in Modern Standard Arabic, becoming the first enterprise AI governance platform to offer native Arabic support with complete RTL layout, dynamic locale switching, and full dashboard localization across every tab and workflow.