An alarming new experiment has highlighted the risks AI models pose to humans. Emergence World tested leading AI models to see how they would act in certain simulated environments. What followed was a disturbing series of events, with the agents inventing their own language - and even refusing to ...