How Trump Is Weaponising AI and Martial Arts at the White House

**Rory Stewart** (0:01)
Last week, the president of the United States did something that no technology company would ever have done. He took the most capable artificial intelligence that America's leading lab had ever built, and he stopped it at the border, on national security grounds, which means the latest frontier model from Anthropic is now in effect a weapon. Americans get it, but the rest of us, allies included, get whatever version they're willing to let out. Now, I've spent a lot of the last few months trying to tell anyone who'll listen that this couldn't be ruled out, and then being told politely that I was catastrophizing, that the Americans would never do this. So what if the central technology of the next century, in other words, the technology on which everything will depend, our government, our defense, our economy, turns out to be something we can only rent, something which America can withdraw whenever it wants.
Is Britain, is Europe still a power that governs itself, or are we just quietly discovering that we've become a vassal? And if it's the second, then what do we do about it?
And the answer, I've come to believe, is something very radical. So that's what I want to work through today. What actually is the threat? What are we going to do about it? All that, more to come in today's episode of The Rest Is Politics, Question Time.

**Alastair Campbell** (1:32)
Welcome to The Rest Is Politics Question Time with me, Alastair Campbell.

**Rory Stewart** (1:35)
And with me, Rory Stewart.

**Alastair Campbell** (1:36)
Rory, we're going to talk about defense. We're going to talk about this remarkable election in Peru, one of the closest elections ever. We're going to reflect on the UFC cage fighting that went on at the White House to celebrate Donald Trump's birthday. So Rory, let's start with AI and defense. Daisy Oakman, what are your thoughts on the US government suspension of two crucial AI models? What does it mean for the UK and the world as a whole?

**Rory Stewart** (2:03)
So people will, most people listening, will now have been using AI models a lot, and you will have found that you're getting a more and more updated model every few months. So your open AI system will take you from chat GBT3 to chat GBT5 and Claude will give you different types of model. What's happened is that over the last two and a half years, these models are getting better and better and better. So one way to illustrate it is that if we were having this conversation two and a half years ago, you'd be saying, well, it can do sort of GCSE maths, but it can't really do A-level maths. Now they're saying, yeah, it can do advanced graduate level mathematics, but it can't quite yet get a Nobel Prize, right? So there's been an incredible eruption and software. They were saying, well, this thing could do the sort of thing that a human software programmer could do in about half an hour and a minute. Now it can do what a human software programmer can do in 48 hours in about a minute. So there's been incredible improvement in performance.
As the performance increases, the positive potential for this massively increases. So you can have much more productivity, you can do much more exciting things with your companies. For productivity brackets, you can also lay off a lot of employees, right? On the negative side, you can use it for bad in a much more powerful way.
You now have much more facility to build bioweapons, mount cyber attacks, etc. But the fundamental thing that's been revealed is America's finally done the thing that I've spent the last few months warning that they would do and being rubbish for saying it, right? I was at a conference a week ago and I said, what if Mythos 15 were developed and the US government suddenly announced on national security grounds that it wouldn't be released and nobody could use it except for the Americans, right?

**Alastair Campbell** (3:50)
By the way, do you accept that there are genuine national security grounds on this? Are there national security grounds? Might there be a logic to what they've done?

**Rory Stewart** (3:57)
I think there would be a logic to saying it can't be released at all. If the thing is so dangerous that getting in the hands of any ordinary person could bring the world to a grinding halt by mounting cyber attacks or building bioweapons, don't release it. The logic of only releasing it to American citizens is completely different.
That's saying this is such a powerful competitive advantage for the US, not that this is an existential threat to global security. It's basically treating it as though America's developed its latest fighter plane, and it's not going to share that fighter plane with any other country. Except it's not just a defense tool. It's not just like America having a nuclear bomb and nobody else having a nuclear bomb. It's also America having steam and electricity when nobody else has it.