Trust at Runtime: Rethinking Identity, Access, and Control for Agentic AI

Agentic AI is fundamentally changing the relationship between identity, access, and enterprise security. Traditional identity models were designed around relatively predictable assumptions: a human authenticates, receives access, and operates within defined permissions. AI agents don’t necessarily behave that way. They can operate continuously, call new tools, assume roles, request credentials, and take actions across systems on behalf of people and other machines.

As organizations move agentic AI from experimentation into production, that creates an entirely new control challenge. Static permissions and periodic access reviews may no longer be enough when autonomous systems make decisions and request access in real time. Leaders need to know not only what an agent did, but whose authority it acted under, why the action was permitted, and whether that authority was still appropriate at the precise moment of execution. HashiCorp identifies this shift from static compliance toward dynamic, runtime-enforced controls as one of the defining security requirements of agentic AI.

Join Blend and HashiCorp, an IBM Company, for an executive dinner exploring how identity, zero trust, and security must evolve for an agentic enterprise. Through candid peer discussion, we’ll examine how organizations can govern non-human identities, establish appropriate boundaries for autonomous systems, preserve the connection between human intent and agent action, and scale AI without sacrificing visibility, accountability, or control.