[rxhampton]

Well, OpenAI and Anthropic will volunteer to get the "too dangerous to release" government stamp before the IPOs.

Or xAI will claim a new dangerous model that will be audited and rubber stamped by "big balls" for the government.

Though the current Trump favorites are hard to discern: The Kushner family has investments in OpenAI and the Trump/Musk relationships are always changing.

With the number of Chinese researchers in the field they cannot keep anything secret anyway. They don't want to either, everyone knows that real military AI does not require the amount of compute that LLMs do and current models are already good enough for automated propaganda.