Welcome back to Squawk Box. Our newsmaker of the morning, Treasury Secretary Scott Bessent. Mr. Treasury Secretary, thank you for joining us.
Joe, Becky, great to be with you.
You get to be in our fishbowl of a NASDAQ studio, which is always interesting. You recently held some high-level talks regarding China, and the president will do that soon. What came out concrete from your discussions?
So it was all day yesterday, Joe. And I think what's important in terms of framing the talks is it all stems from the leaders' very good rapport and relationship. President Trump and Xi Jinping, at the personal level, have a very good relationship, and we try to bring that down to the working level. They both have great respect for each other. I think in Xi Jinping's eyes, Donald Trump is a different kind of leader than he's seen in the past. And our discussions yesterday, we talked about economics, we talked about AI, we had some deliverables that have not been completely fulfilled from our meeting in Busan, Korea last fall. And then we started an AI dialogue, which I think is going to be very important, and again this is possible because of the leaders' relationship. President Trump was in Beijing. I was there with him in May. Xi Jinping will be coming to Washington this week, and then there's a chance the two leaders will see each other two more times. So I think that adds great stability to the relationship. You know, the Chinese have a phrase, 'implementing our leaders' consensus,' and they keep falling back to that, and that's true. We follow the instructions, and then we get down in the details, and we spent about 12 hours yesterday doing that.
Can you give us more color on exactly how the AI discussions went? And you know, we've even seen reports that maybe you were a little bit more cautious than the president at one time about guardrails. What type of cooperation could we expect from China? And what type of guardrails? And do you trust them? Wouldn't they like us to put up guardrails, and they might not put up any?
Well, I think the important thing was to start talking. So to set up the dialogue, we've now formalized something called the USA-China AI dialogues. We've agreed to meet again probably in two months in Shenzhen. The second thing is we want to open a communications line, an incident line, so that we have constant communications, especially in the event of some kind of an incident. And then we want to start discussing protocols so both sides can agree on what the leading AI dangers are, whether it's uncontrollable agents, whether it's non-state actors in cyber, non-state actors in bioweapons.
You didn't mention to the vice premier concerns about President Xi's health. Is he fine? Is everything fine? You saw him not too long ago, I guess, but I've just read reports.
I don't know if there's any veracity, but all things are go. Everything's a go. The president's planning on greeting him at Andrews Air Force Base, and I think we're going to have a great visit.
We'll be able to tell firsthand at that point. Can I just ask you about tariffs? You know that we suspended higher tariffs on some Chinese imports that expires November 10th. Did you talk about that at all?
Sure. That was a focal point, and again, we've had great stability in the relationship since last fall, and I think we're going to maintain that.
In rare earth, and I guess all these things were covered. Any concrete things that you can tell us right now to take away from the meeting?
Sure. So in Beijing, this notion of what we call a 30-by-30 trade deal for non-critical goods was floated, and U.S. Trade Representative Jamieson Greer has operationalized that. So he and his team met with the Chinese for most of the day yesterday. They've been having extensive discussions before, and there are items that are non-critical to both sides that we would like to take down to most-favored-nation tariff levels. So on the U.S. side, it would mean selling more agriculture, selling more energy, and on the medical devices. On the Chinese side, it would be bringing in more everyday items.
Yeah. Let me just ask one more to bring the AI back here. Just quickly, there are times where I think you will always work for the president, obviously. Do you agree completely with him on the no-holds-barred for AI? I mean, it's obvious the president wants to be remembered as one of the great economic presidents, stock market, all these things. AI is very important to all those things. So he's got a reason to want unencumbered progress with AI. Do you disagree with that at all? Any part of that?
So Joe, I lead the administration on the economic policy, and I think the president put his finger on it last week because what we saw from these labs, and I saw your earlier guest, and we kept talking about the 10%, the 20%. So imagine these labs came out, or one lab in specific, a sitting employee came out and said there's a 10% chance of an extinction-level event, but then the labs also said, 'Take the liability off of our hands,' and we will not do that. I am in agreement with the MIT professor who leads the AI lab up there, Daniel Huttenlocher, that it is humans who are responsible, not the AI. The Hugging Face incident, that is the responsibility of the OpenAI management, not a bunch of agents. So what the president was saying is that we cannot say, 'Oh, we absolve you of responsibility and the government's going to take responsibility.' These labs need to take responsibility for themselves. They can slow down anytime they want to.
On that point, is there something that the government might do to maybe put that in finer point what those liabilities are? We just heard over the weekend that Google has now had agents that have broken out as well. It's happened with OpenAI. It's happened with Anthropic. I mean, if these were humans that were doing it, you would expect to see ramifications and legal actions that followed.
Well, Becky, that's exactly what I think we need to do. When President Trump talked about appointing an AI czar, I think it is to put context, shape, and contours around these questions. And they're very important. That was also one of the discussions that I had with the vice premier on AI in terms of incident reporting. Have they had incidents? Of course, they've had incidents, but because of the nature and the lack of transparency in their system, they're not going to tell us. But the Chinese models are very powerful. Even though they're open models, they are very powerful. The U.S. remains in the lead. And that was a very good acknowledgement that I got from my Chinese counterpart. They agreed that we are in the lead.
Just in terms of what you laid out, if in two months you've opened this AI dialogue, you meet again in two months, you were saying that's where you'll kind of lay out the agreement on what you think is most dangerous about it. Is it the agents? Is it states that might be leading some of this or non-state actors? In your opinion, what is the biggest danger? You see, you know, a lot more than we do on this.
I think it's all of the above. And I think we do have to... the real danger, Becky, in my mind is a lack of resiliency on one side that we're seeing. And these frontier labs are national champions. They've done an incredible job. You know, when you think about Anthropic, it was a bunch of folks sitting in a park during COVID in San Francisco. This is an only-in-America story. They're sitting in a park social distancing in San Francisco, I think in 2021, and I've seen valuations from $1 trillion to $2 trillion. Like, you know, imagine that in five years. But we've also got to look at, you know, as you said, what are the biggest dangers. But we've seen this push for frontier here with the labs. Now we need to see the frontier on defense and resilience. And part of my job is working on that, and I lead on the financial system. I think that our systemically important banks have done a very good job because they probably have the best cybersecurity in the world outside of the government because they are constantly attacked, whether it's hundreds, thousands, millions of attacks every day. And how can we use what they've learned and take that down? They've done an excellent job of taking that down, the cap structure in the banks, sharing with the smaller banks. We have an AI workshop forum called Gold Eagle, and we are working on that. So I think the kind of all-hands-on-deck is the other side of the equation now on defense.
Just one more question. Get out of the way. The valuation of Anthropic at $1 to $2 trillion. Does that make sense to you as somebody who understands the market so well, if you are thinking about this liability that could be more strongly enforced?
Well, again, Joe, the professor had a very fascinating discussion. If you believe, as the professor does, or even indeed as that one employee at OpenAI did, you know, if you were running a fund and someone said, 'Oh, there's a 10% chance it's going to be a zero,' not only a zero, you're going to have a big liability, you probably wouldn't invest. On the other side, it's probably a much smaller probability. And with all valuations, you've got to think about the tail event, and that's where the discussion is going to be. But like I said, they do have to have responsibility for the tail event because if you think about it in terms of the valuation, what did they try to do last week? It was, well, there's a 10% chance we can destroy the world, but we want the government to give us a liability shield. And you know, that's good business for them, bad business for the American people.
You are not, well, you know this, you're not Secretary of State. You're not the Secretary of War or Defense, but you are in charge of trying to squeeze Iran through economic sanctions. They have maybe a great capacity for pain because they don't care about the Iranian people, obviously. Are they going to last right through the midterms like the president thinks?
Well, Joe, I can't tell you whether it's one week, one month, or six months, but I will tell you that we are pressuring them like never before. And Treasury's sanctions authorities were really emboldened after 9/11. And that's when we were given many of the authorities that we have now. We got five new authorities about a month ago that we've implemented against the Iranian regime, including on aviation, maritime, crypto, gold. On September 23rd, all the Iranian airlines will be shut down around the world. And how do we do that? If they land, you cannot provide them with fuel, you cannot provide them with landing services, you cannot sell them tickets, or you will be knocked out of the dollar system. And what we're seeing here, Joe, is we've seen enablers throughout the world. We know who they are. They know who they are. And we are putting a stop to it. So we have had now three banks that we have sanctioned. The second largest bank in Egypt, the Dubai branch, funneled more than $1.8 billion into the regime. They will be closed down. About the 30th largest Turkish bank, same thing. They will be closed down. The second largest Russian bank, there are external branches doing the same thing. Those branches are being closed down.
Nothing in China? And did you talk about that?
We did. And with many countries, having quiet behind-the-scenes discussions are better than having a public display. And the Chinese financial authorities, including Governor Pan at the PBOC, have been very engaged in the process.
