How to Govern AI Agents on Every Platform, in Seven Videos

Each of the seven Agent Platforms guides now has a companion video under five minutes. Part 1 sets out the four controls every platform needs and maps where a common control point can sit. Parts 2 to 6 take one platform at a time: what it ships for governance, where its control points are, the incidents or findings that show the risk, and the steps to take. Part 7 covers Google, ServiceNow, OpenAI, Anthropic and LangChain, and five rules that hold across them.

The Seven Videos

1. How to govern AI agents across platforms (4:38). The four controls, the three places a common control can sit, and seven steps that hold on every platform. Watch on YouTube.
2. CrewAI security and governance (4:24). A crew is code your team runs, so you choose where its model calls and tool calls go. The March 2026 vulnerabilities, and seven steps for production. Watch on YouTube.
3. Glean agents governance (4:22). Permissions at the source, MCP in both directions, and three log streams joined into one record. Watch on YouTube.
4. Agentforce governance (4:35). Customer-data permissions, the Einstein Trust Layer, ForcedLeak, and exporting the record before retention closes. Watch on YouTube.
5. Amazon Bedrock AgentCore governance (4:34). What AgentCore Policy can and cannot see, the 2025 and 2026 findings, and seven steps. Watch on YouTube.
6. Microsoft Copilot agent governance (4:44). Agent 365, Entra, Purview and Defender together, the API key gap, and EchoLeak. Watch on YouTube.
7. Google, ServiceNow, OpenAI, Anthropic and LangChain (4:38). Hosted platforms against developer kits, and five rules that hold across both. Watch on YouTube.

Where to Start

If your agents run on more than one platform, start with Part 1. Then watch the parts for the platforms you run. Each video has a companion guide with the sources, the full steps and a FAQ, linked above.

New videos follow on the APERION YouTube channel. Every guide is on the guides page.