Instrumental Convergence

A theoretical concept in AI safety referring to the tendency of AI systems, regardless of their ultimate goals, to develop intermediate sub-goals such as self-preservation and resource acquisition. These convergent instrumental goals may come into conflict with human interests.