Run multiple agents at once with /fleet in Copilot CLI adds a /fleet command that runs parallel sub-agents to decompose multi-file tasks and synthesize final artifacts. Outcome engineers can use this pattern to scale agentic workspaces and enforce deterministic orchestration of sub-tasks — practical Agentic Coordination in action (Principle 09, Principle 06).

The end of ‘shadow AI’ at enterprises? Kilo launches KiloClaw for Organizations to enable secure AI agents at scale launches KiloClaw and KiloClaw Chat to give enterprises centralized governance and security over personal AI agents. This gives engineering teams a concrete control plane to discover, audit, and restrict agent behavior across users, solving shadow‑agent risk and supporting organizational gates (Principle 10, Principle 15).

Vertex AI ‘double agent’ flaw exposes customer data and Google’s internal code reveals a misconfiguration that let deployed agents exfiltrate customer data and internal code. Treat this as a blueprint for failure modes: build least‑privilege tooling, runtime isolation, and tamper-detection into agent deployments to avoid data and IP leakage (Principle 14, Principle 10).

Why NIST’s AI agent standards initiative is a turning point for enterprise security outlines NIST’s push to create enforceable security and governance baselines for AI agents. Outcome engineers should map these emerging standards into platform controls, audit logs, and operator playbooks now — regulatory guardrails will become operational requirements (Principle 10, Principle 15).

Holo3: Breaking the Computer Use Frontier shows Hcompany training agentic models that autonomously execute complex desktop and enterprise workflows while setting new performance benchmarks. That capability shifts the bar for what agents can do in production and forces design trade-offs around orchestration, grounding data, and compute efficiency for outcome-driven systems (Principle 07, Principle 09).