I would laugh my ass off if Coca Cola Company ends up being the company that solves alignment - so that it can align an "open weight" AI with its corporate interests.

Without that though? Our ability to manipulate LLMs is so shaky I would be really surprised if anyone managed to pull off this kind of model manipulation and have it remain undetected.