Kam Sween is a software engineer specializing in secure AI-agent authorization: ensuring autonomous applications act for identifiable users, operate within defined permissions, and obtain consent before consequential decisions.
His background includes more than a decade building secure, compliant cloud platforms and modernizing regulation-heavy legacy systems. At AI Engineer World’s Fair 2025, he was a staff engineer at Auth0, an Okta company, and technical lead for its AI Frameworks & Services team, developing frameworks and SDKs for AI applications.
Putting human consent inside agent workflows
In a joint conference session with Bobby Tiernay, Sween demonstrated the architecture of a simulated TypeScript trading assistant that identifies its user before requesting authorization for a stock purchase. His approach centers on three practical safeguards:
- User-bound agent identity: OpenID Connect and device-based authentication associate each agent session with a specific person, making actions attributable without relying on shared service credentials.
- Auth0 Token Vault: Sensitive authorization tokens remain separate from application code, allowing agents to access resources on a user’s behalf without embedding secrets.
- Client-Initiated Backchannel Authentication: Before executing a sensitive action, the agent requests approval on a trusted device, identifying both the agent and the proposed transaction. Sween used a blocking callback for simplicity while noting that production implementations can use asynchronous approval or polling.
When unreliable connectivity interrupted the demonstration, Sween explained the intended authorization sequence without claiming the simulated trade had succeeded. His focus is making identity, consent, and accountability foundational properties of agent software.
