AI systems need partitioned threat models, not a single risk approach.

In a Help Net Security interview, Naor Penso, CISO at Cerebras Systems, explains how to threat model modern AI stacks. He emphasizes avoiding a single risk approach, advocating for partitioning AI systems by function and impact. This method is crucial for effective security. Penso also discusses framing threat modeling for business leaders. He highlights how assumptions change as AI integrates into core infrastructure. The interview explores optimal strategies to partition AI stacks for threat modeling, ensuring teams avoid oversimplification.