In her keynote at the Software Architecture Gathering 2025, Cheryl Hung – tech executive, cloud-native expert, and former leader at the Cloud Native Computing Foundation – explores whether AI coding assistants are capable of building secure authentication systems. Through a practical experiment, she demonstrates how quickly AI can generate working solutions, but also reveals the hidden security flaws, architectural challenges, and organizational questions that emerge beneath the surface.

Key Takeaways:

1. Authentication becomes even more complex in a future with AI agents and federated identities
2. AI-generated code can appear correct while hiding critical security vulnerabilities
3. Security remains as challenging today as it was five years ago despite significant industry investment
4. Misconfigurations are often a bigger threat than sophisticated attacks
5. "Shift-Down Security" moves common security concerns into platform teams through automation and standardization

Watch the full keynote to learn why secure identity management remains one of the hardest problems in software architecture and how organizations can better distribute security responsibilities across teams.