“Leopold Aschenbrenner wrote and shared an internal memo with OpenAI board members asserting that the company's security was insufficient to prevent the theft of model weights or key algorithmic secrets by foreign actors.”