“The primary obstacle to a binding international AI safety agreement is the lack of trust between major powers like the United States and China, with both sides concerned the other would secretly cheat on the deal.”