A study shows major artificial intelligence models are likely to refuse to criticize restrictive leaders worldwide. The Meta ...