Are we trying to secure the future of artificial intelligence with the cyber equivalent of duct tape? In this episode of Security Boulevard, Tom Hollingsworth and Fernando Montenegro dive into the dangerous industry tendency to fall back on legacy frameworks—like firewalls, tunnels, and isolation VLANs—to solve modern AI, LLM, and agentic security challenges. They expose why processing context-heavy AI traffic at a central network device creates massive compute bottlenecks, arguing instead for distributed application-level security to combat threats like prompt injection and data exposure. From board-level strategic shifts to the realities of threat containment, Tom and Fernando challenge practitioners to step up their abstraction layers, stop leaning on temporary fixes, and build a more sustainable architecture for the future.