Researchers in England have concluded that perfect alignment between AI systems and human interests is mathematically impossible, shifting the focus from elimination of misalignment to its management.