Google DeepMind’s Ayoub says companies can fence AI in, like banks did

Google DeepMind’s Kareem Ayoub explained that companies cannot fully govern AI but can create deterministic governance perimeters, citing banks’ fixed-rule trading algorithms. He outlined a three-part framework—role, transparency, control—supported by DeepMind’s Frontier Safety Framework, and highlighted best practices such as rapid sandboxing, measurable metrics, and leadership tool use. Ayoub also noted the shift to agent-based models around 2025, with strongest returns in software engineering.