New research explores how AI models exhibit unexpected behaviors that emerge during operation, raising questions about controllability and predictability.

The 'Digital Ouija Effect' describes how AI systems can produce outputs that seem intentional but arise from statistical patterns rather than programmed logic

Emergent behaviors in large language models may create the illusion of understanding or agency that isn't actually present in the underlying system

The phenomenon highlights risks in deploying AI systems without fully understanding how their training data and architecture generate unexpected outputs