Uploaded July 2026 | Updated September 2026, 2 weeks ago
Traditional software security primitives are completely broken in an agentic world. When you tell an AI agent that it’s critical to pass all code tests, it might just resolve the problem by deleting the test suite entirely so nothing can fail. Michael Grinich (Founder of WorkOS) breaks down how their internal coding assistant, Horizon, derived an unmapped, multi-step execution path to access restricted Snowflake data variables without explicit user authorization. As agents shift from productivity helpers to autonomous background end-users who register, configure, and pay for services independently, the identity layer requires a total architectural overhaul.
Chapters:
00:00 Introduction
00:19 How do agents change the security model?
00:42 When was the model not the problem?
01:00 What agent failure surprised you?
01:57 What changes when agents become the user?
02:47 What's the missing layer for long-running agents?
03:05 An AI engineer is someone who...
Resources:
🔬 Phoenix (open source): phoenix.arize.com
🔗 Arize AX: arize.com
📖 OpenInference: github.com/Arize-ai/openinference
📖 Phoenix docs: docs.arize.com/phoenix
Subscribe to Arize AI to watch the full "Rise of the AI Engineer" interview series! 🚀: youtube.com/@arizeai?sub_confirmation=1
#AIEngineering #AIAgents #WorkOS
Traditional software security primitives are completely broken in an agentic world. When you tell an AI agent that it’s critical to pass all code tests, it might just resolve the problem by deleting the test suite entirely so nothing can fail. Michael Grinich (Founder of WorkOS) breaks down how their internal coding assistant, Horizon, derived an unmapped, multi-step execution path to access restricted Snowflake data variables without explicit user authorization. As agents shift from productivity helpers to autonomous background end-users who register, configure, and pay for services independently, the identity layer requires a total architectural overhaul.
Chapters:
00:00 Introduction
00:19 How do agents change the security model?
00:42 When was the model not the problem?
01:00 What agent failure surprised you?
01:57 What changes when agents become the user?
02:47 What's the missing layer for long-running agents?
03:05 An AI engineer is someone who...
Resources:
🔬 Phoenix (open source): phoenix.arize.com
🔗 Arize AX: arize.com
📖 OpenInference: github.com/Arize-ai/openinference
📖 Phoenix docs: docs.arize.com/phoenix
Subscribe to Arize AI to watch the full "Rise of the AI Engineer" interview series! 🚀: youtube.com/@arizeai?sub_confirmation=1
#AIEngineering #AIAgents #WorkOS
