Are there any reasons to believe that LLMs are in any way more alignable than other approaches?

So, ideally you would like to assume only

1. □A→B
2. □B→A

and conclude A and B ?