Arthur launches enterprise agent governance platform: control

On Dec. 17, Arthur released an Agent Discovery & Governance (ADG) platform. It inventories agents across compute environments, applies continuous policy enforcement, and integrates with major ecosystems including Google Cloud Gemini, AWS Bedrock, and Microsoft Agent Foundry (Arthur announcement).

On Dec. 18, Automation Anywhere reported Q3 results: 45% year-over-year growth in AI bookings; AI bookings now exceed 70% of total bookings; and remaining performance obligations (RPO) rose more than 20% year over year. The vendor also highlighted new multi-agent orchestration in its Mozart Orchestrator (Automation Anywhere press release).

Kyndryl on Dec. 16 called for a rebuilt security blueprint for the agentic era. It warned that prompt injection, data poisoning, and multi-agent chain reactions can defeat static defenses and require DevSecOps and adaptive monitoring (Kyndryl article).

Technically, vendors are shipping three complementary capabilities: discovery (automated agent inventory), orchestration (multi-agent coordination and tool calling), and governance (policy, acceptable-use controls, and observability). McKinsey research cited by vendors and security teams reports roughly 80% of organizations already seeing risky behaviors from AI agents, underscoring the operational gap enterprises face (McKinsey, Oct. 16, 2025). Automation Anywhere also flagged ISO 42001 certification activity and product updates that add enterprise-grade policy and rollout controls.

Implication: expect more platform integrations, governance modules (policy engines, memory controls, RBAC), and security services focused on continuous validation and incident containment as organizations scale agent fleets.

Why It Matters

• Inventory first: automated agent discovery is table stakes—if you can’t find agents running in your environment, you can’t govern them. Prioritize discovery across cloud and edge.
• Add continuous validation: build observability and policy gates into agent workflows (task-level KPIs, data-access audits, anomalous tool-call detection) to surface drift and exposures.
• Orchestration reduces blast radius: use multi-agent frameworks to enforce sequencing, rate limits, and tool-call constraints to contain systemic failures.
• Security by design: adopt DevSecOps for agents. Integrate prompt-safety checks, data-handling policies, RBAC, and adaptive monitoring into CI/CD and deployment pipelines.