Full Workshop: Better Auth — Paola Estefania, Better Auth

The Agent Security Protocol proposes shifting AI agent authentication from shared user credentials to distinct agent identities, enabling granular authorization and full auditability 1:56.

Key Takeaways:
• Current practices often grant agents full user credentials, effectively letting them "pretend" to be the user, which poses significant security risks 1:59.
• The protocol introduces a "phone book" directory where agents can discover available service capabilities and map user intents to specific actions 5:14.
• Authorization is based on specific capabilities rather than broad scopes, allowing users to grant precise permissions like read-only access 6:04.
• Each agent receives a unique identity with a private key, enabling traceability and the ability to revoke access to specific agents without affecting others 7:58.
• The system supports both host-based and user-based policies to balance security with user experience, minimizing unnecessary approval prompts 32:21.

This approach transforms AI agents from hidden proxies into accountable, principal actors with their own lifecycle and audit trails.

Sources:

• 1:56 Introduction of the problem: agents acting with user credentials.
• 5:14 Concept of a directory for agent capability discovery.
• 6:04 Shift from broad scopes to fine-grained capabilities.
• 7:58 Importance of unique agent identity for traceability.
• 32:21 Balancing