Tom and Penny explore AI kill switches, rogue agents and the growing argument that AI should now be treated as critical infrastructure
This week we - along with the US and UK governments, as well as OpenAI itself - are tackling the controversial topic of AI kill switches. Tom and Penny discuss the proposals on the table, as well as the practical challenges of defining what a kill switch actually is, who would control it and what circumstances would justify its use.
We also examines the wider implications of AI's transition from software tool to business infrastructure. From reports of rogue agents to the challenge of incorporating AI outages into resilience planning, we ask whether organisations are becoming dependent on systems whose governance remains unresolved.