Securing the AI Control Plane with Speakeasy
Summary
Host Rich Trafalino interviews Speakeasy CEO Sagar Baciu, alongside panelists George Finney and Nick Espinosa, on securing the AI control plane. They discuss challenges like non-human identities, data governance, and the balance between oversight and surveillance. The episode covers Speakeasy’s enterprise solution for auditing agents, preventing prompt injection, and ensuring compliance while keeping humans in the loop for critical decisions.
Topics discussed
Introduction: The challenge of securing AI agents and identity
Speakeasy's AI control plane: Value, capabilities, and pricing
Defining agent permissions and establishing agent identity
Scope of governance: From workforce AI to comprehensive coverage
Balancing governance with employee privacy and surveillance concerns
Data privacy: Preventing unauthorized training on corporate data
Human-in-the-loop: Decisions that must remain with humans
Case study: Could Speakeasy have stopped the OpenAI incident?
Accountability: Who is responsible for autonomous agent actions?
Implementation: Required organizational sophistication and resources
Avoiding cognitive surrender: Keeping human judgment in the loop
Current enterprise adoption and closing remarks
Listen ad-free on Castria