Bobby Tiernay

Bobby Tiernay is a principal architect at Auth0 and Okta focused on making AI agents accountable to the people they represent. His work applies identity standards, fine-grained authorization, and user consent to software capable of retrieving private information, calling external services, and taking consequential actions.

His earlier technical work included Java, Elasticsearch, and distributed-data tooling; a public code example also connects Hadoop filesystem handling with Apache Spark. After years working across Auth0 and Okta’s identity platforms, he expanded from access-control architecture into AI-agent security. His projects have included Auth for Gen AI, Auth0’s Model Context Protocol server, and Tenant Security Manager.

• User-bound agent authorization: Replace shared API keys and overprivileged service accounts with short-lived OAuth tokens tied to a specific person, service, and operation. Token exchange preserves identity, limits permissions, and makes actions auditable.

• Retrieval-layer authorization: Enforce document permissions before sensitive material enters a model’s context. Fine-grained access controls prevent retrieval-augmented systems from exposing information simply because an agent can reach it.

• Human approval for consequential actions: Use Client-Initiated Backchannel Authentication to request consent on a trusted device before an unattended agent completes a sensitive operation. At AI Engineer World’s Fair 2025, collaborator Kam Sween illustrated this approach with a stock-trading workflow.

• Secure Model Context Protocol infrastructure: Apply delegated authorization and trustworthy client identity as MCP moves from local tools to shared remote servers. Tiernay also contributed to the protocol community’s OAuth client-registration discussions.

• Specification-driven development: Favor explicit requirements, sandboxed experimentation, and progressive security controls when AI generates software. His perspective on security by design treats precise communication of intent as an increasingly important engineering skill.