What Happens When AI Becomes Critical Infrastructure?

AI is moving into critical systems, business operations and infrastructure where simply shutting it down could cause more harm than leaving it running. The challenge is no longer just how we control AI, but how we isolate, govern and remain resilient around systems we increasingly depend on. In his third BASELINE conversation, Gavin Ray explores what happens when autonomous agents are given objectives and access to resources without the sanctions that govern human behaviour. They discuss why the traditional AI “kill switch” may already be the wrong idea, the widening gap between technological capability and governance, the extraordinary capital flowing into AI infrastructure, and why the final 5% of creating real commercial value remains stubbornly difficult. The conversation also exposes a less visible consequence of enterprise adoption: as AI enters organisations and accumulates their knowledge, processes and experience, are companies adopting the model, or gradually distilling themselves into it? And if decades of human expertise can eventually be queried on demand, what becomes of the value of time and talent?