MCP 2.0 introduces three foundational security controls that didn’t exist in the original specification—and they directly address how AI agents execute work inside your enterprise.
In this clip from the “MCP 2.0 Explained: Securing AI Agents in the Enterprise” STRIVE podcast, Werner Nel, Principal Security & AI Intelligence at Commvault, breaks down the technical changes: OAuth for least privilege access, structured schemas to block prompt injection attacks, and elicitation flows that let you pause AI workflows for human approval on high-risk actions.