Skip to content

BRK250 Breakout Thu 19 Nov 11:30 PST

Govern AI everywhere: From open frameworks to enterprise control

Mohamed Elmergawi, Mehrnoosh Sameki, Manmeet Dhhillon, José Mejias

Summary

From the recording

Microsoft announced the launch of Agent Controls, a centralised policy decision point for agents across runtimes and clouds. It builds on open-source Agent Hooks and the Agent Control Spec, and is based on Azure content safety systems. It is offered as a standalone service and built into Microsoft Foundry and Agent 365. A third open-source tool, Assert, creates test scorecards. In the demo, a 37.5% policy violation rate fell to 0%. Caveat: no GA or preview status, pricing or licensing was stated; the result came from a single demo.

Session description

Microsoft's catalogue description

Your teams already picked their tools, and telling them to start over is how an AI program loses a year. This session shows the alternative. Start with open source that runs against agents built in any framework, then bring those same agents into Microsoft Foundry when you need enterprise identity, policy, and one view of quality across the estate. Leave knowing what you get for free, what requires a platform, and where lock-in risk actually sits.

← All sessions