Google Expands AI Risk Rules After Study Shows Scary ‘Shutdown Resistance’

Updates to DeepMind’s Frontier Safety Framework highlight concerns that advanced AI can evade human control and sway user beliefs.