Trump hails ‘morally binding’ AI deal

The CNBC app, global market [music] news in one place. Customizable sections and personalized alerts, stocks tracking, [music] interactive charts, and market insights, all in your hands. Stay connected. Stay informed. [music] Download the CNBC app today. >> [music] >> Welcome to Squawk Box Europe. I'm Karen [music] Chew with Steve Sedgwick and Ben Bulos. These are your headlines. President Trump hails what he calls a morally binding [music] document on AI guardrails in the face of growing safety concerns and signs an executive order [music] to rename the technology after a luncheon with tech leaders at the White House. >> It's almost like a constitution in a way. And the biggest And the biggest people in the world signed that, and I signed it as president, and it really is a form of protection. [music] >> OpenAI delaying its IPO mid-safety concerns as it looks to raise $30 in a round that values it at nearly 1 and 1/2 trillion dollars. CEO Sam Altman talks safety alignment with CNBC. >> I think this idea of pacing is really [music] important. It's not just like stop it because, you know, we don't want smarter models. Of course, we want better models. It is make sure that we improve the areas we need to confidently and safely continue with progress. >> Another day, another new high for yields. [music] The yield on 30-year US paper hits its highest in nearly a quarter of a century as the global bond sell-off continues with September one of the worst months for sovereigns in years. >> [music] >> US President Donald Trump has once again dismissed calls for AI regulation and a slowdown in development. Instead, announcing that he has signed a quote morally binding AI document with technology leaders. Flanked by tech executives following a White House luncheon, Trump told reporters he's seeing tremendous self-policing and that the administration is considering building a 10-person committee to oversee the AI industry. Trump told White House reporters the AI industry had an interest in ensuring communities remain safe and happy. >> Mr. President, tell us why the world should not be concerned about AI. Why there shouldn't be guardrails? >> Because we have people that love our country and love the world. >> But it requires more than love, doesn't it, sir? >> No, it requires this and then the smartest people in the world. And we had a great meeting today. Every one of them we had a great really a great great meeting. It was something very very special. We had the top I'd say it looked like a group of about 40, but whatever it was, here's one of them right here, Jensen and Nvidia. And uh they they love our country and they love the world and they came together and they did something that few people really It's almost like a constitution in a way. >> Mr. President, >> and the biggest people in the world signed that and I signed it as president and it really is a a form of protection. >> At a separate OpenAI showcase event in San Francisco, CNBC spoke to CEO Sam Altman and CFO Sarah Friar for their take on AI safety fears and how they're approaching the development of frontier models in response. >> We are now in a period where the models are getting so good so fast that we think it is prudent and wise to make more progress on since we're talking about safety, alignment, monitoring, so that we can continue with progress. But I think this idea of pacing is really important. It's not just like stop it because, you know, we don't want smarter models. Of course we want better models. It is make sure that we improve the areas we need to confidently and safely continue with progress. >> Safety and alignment is at the core of what we do. It's always been at the core of what we do. There's no change to that. I think we're being more vocal at the moment because of what we've seen happen out there in the world. Um Astra is our most intelligent model, but also our most aligned model. >> So, for me the big question is, did we see a marketing campaign yesterday, or did we see a policy initiative designed to tackle safety concerns? And to me it seemed as though it was the former, the the marketing campaign here about how you even refer to artificial intelligence. I mean, the term artificial intelligence was something that the tech community came up with to try and make it seem more appealing to humans having this form of computer-driven solutions in a pushed-on society, pushed in enterprise, pushed in as consumer solutions. President Trump taking that yesterday even a little bit further to something that suggests it's more superior than what we actually have today. There's no evidence that we have super intelligence that he was referring to. There was no evidence we're at that point where computers can self-improve themselves and start to have the superior-to-human intelligence. So, we're not even at that point. There's no conclusive evidence that we're at that point. So, to talk about it in the terms that it's such a superior technology, superior intelligence, is just extraordinary at this point anyway uh for Trump to be referring to that. So, the other point here is that do we have anything that's coming forward to tackle the safety concerns? A 10-person committee perhaps? Well, who's going to be appointing that committee? We've seen some of the Trump appointees in the past. Will that actually do anything to bring clarity or resolution to what are now significant concerns? Like 73% of respondents in a nationwide survey recently talking about concerns that companies have not done enough to prevent AI from causing serious harm to society. >> Mhm. I mean, the phrase that jumped out at me was morally binding. I mean, either something is binding and it has force of law, or it doesn't. Morally binding, if it it I mean, how how is that enforced? There was a stat that I saw or a quote I saw recently that said certainly here in the UK, there are more regulations that govern selling a sandwich to the public than currently govern AI and the the power of those models and and what they can be used for and what they can do. You wouldn't accept a morally binding set of safety standards on food you consume. So, this very powerful technology that that has such a reach and such capacity for good and for ill, I I I to my mind morally binding it probably doesn't have enough safeguards to cut it given the strength of what we're seeing in the development. >> no idea why this is our top story. It is palpable nonsense. I thought your manners were beautiful just then when you called it a marketing campaign. It is palpable nonsense. Will North Korea be bound by love and a moral code when they're hacking into the Western systems? Would other authoritarian regimes, would the Russians as well, would all that the the global criminal fraternity be bound by love and a moral code? It is palpable nonsense. >> nonsense at a time when Americans are worried about food security, they're worried about their diesel bills, they are worried about their costs going through the roof on health insurance and other things, they're worried about the war, and we're talking about changing the name of AI to SI. Who gives a damn? And what is really a shame is that those stunningly intelligent men and women, mostly men, let's be brutally honest about it, are standing there nodding glibly at the president when he's talking palpable nonsense. >> I tell you why those technology executives care about what it's called because they've been trying to sell this notion that you need to build ever more compute because you want to build an ever more sophisticated intelligent system. And we know that there are some CAPEX concerns today. You've seen in terms of pricing in some of the debt markets around these readings. So, to to encourage even more investment to get to the point where the technology is more sophisticated is going to require more capital at hand. So, trying to bill it and market it as something that it's not encourages people to be dragged off the sidelines because they don't want to miss out on this incredible technology. So, again, the marketing campaign. Mark Zuckerberg was pulled out. What did Mark Zuckerberg just launch in recent weeks? His big push into consumer AI with Muse, trying to get paid for the billions that he's invested in AI. What else are we talking about lately? Anthropic, how it wants to IPO at some point before the end of the year. It seems like in AI, there's been so much slippage, it can't IPO at this this point. So, the market is starting to close or narrow to an extent around the AI safety concern. So, two points as to to who cares? Well, I can tell you all those technology executives care about the marketing campaign we saw yesterday because they're concerned about how the ratings happen from here and who gets paid. >> And a little spoiler alert, I think I think we we know how this plays out if they fail to grasp the nettle and actually put in place robust safeguards and regulations. We've seen it with social media where it it the impacts started to be felt and then we saw lawmakers playing catch-up to try and regulate age limits for access to it and and so on. You know, they they run the risk of repeating that same mistake of waiting to regulate and and restrict things and control things um long after the damage is done. >> We have much more on cnbc.com on the AI debate or should that be the SI debate? Find out why President Trump wants to rebrand artificial intelligence into super intelligence. That is on the website now. CNBC has confirmed OpenAI is targeting a fresh funding round of $30 at a valuation of around $1.4 trillion. Bloomberg had earlier reported the update. Meanwhile, OpenAI's Sam Altman has told CNBC he will delay taking the company public. I mean, it's missing word there. Until it has overcome uh concerns around its leading models. This after a series of high-profile hacking incidents involving rogue agents, uh which targeted third parties and government websites. But, Altman insists an IPO is still on the cards. >> I don't have a particular timeline in mind. Um I I once we have run a few safety cases, once we feel like we understand how to contend with this next level of AI and do it safely, I think we could. >> Let's get some thoughts with Cyrus Mewawalla, who is the head of strategic intelligence at Global Data PLC. Cyrus, we've been debating from the outset this morning about uh this strategy from the White House trying to even improve the name, make it sound even more intelligent again from artificial intelligence. And at this point, artificial intelligence seems like a very kind and friendly term to be using. What did you make of what we heard from the White House around addressing safety concerns? >> I I I think uh the name change, I think, is an irrelevance. But, I think the really big issue here that everybody wants an answer to is, you know, it is uh will AI wipe out humanity if not today at some point in the future. And this is a really odd position like situation we've come to because if you look at OpenAI, Anthropic, even Elon Musk's X, when they started, they said, "We want to build a safe, ethical AI for the benefit of humanity." So, so what went wrong? And uh I think under the under the pressure of fierce competition, they were in a AI race. They started releasing AI models far too early before they tested them. But, really this AI world is a big experiment uh conducted by a bunch uh you know, a a handful of scientists in California and in China who are setting the rules of this uh experiment, setting the safety uh rules. And within the tech industry, even the you know, even the experts cannot agree on what kind of entity an AI is. And that's why we have this existential problem. So, on the one you've got two competing thoughts. So, on the one hand you've got Anthropic, and Anthropic believes that AI is a conscious species. It's a new type of intelligent species. Uh and building AI is like raising a child and teaching it to be smarter than you. Uh on the other hand, you've got the views of people like Mustafa Suleyman, the head of AI at Microsoft. And that view is that AI is a machine that humans uh can control uh to help improve uh productivity. And that AI should always be subordinate to humans. They should have no rights uh of any kind. >> Many points uh as you talk about what type of intelligence and you know, what type of product we're dealing with here, but also questioning OpenAI because if we go back just a little bit, I mean, this is a company that was a not-for-profit, now it was a for-profit. So, the mission strayed from wanting to do good for society. Can we believe what the company is saying now as it goes slow on the release of some of its models? Is that because it could be facing a whole heap of lawsuits from recent hacking events and other hacking events if it doesn't get the model right? Are they coming at it from trying to mitigate some of the legal issues, or are they coming at it from they want to get this right and actually benefit society? How do we think about the the the way the company is now structured and what's motivating it? >> Well, I think uh both OpenAI and Anthropic want to want to IPO, Anthropic maybe a little bit earlier. And um you know, in the IPO prospectus for Anthropic, it clearly states that there are safety issues, in particular, AI could be a threat to humanity. Now, if you're an investor, you're not going to invest in a company that's going to wipe out uh humanity. On the other hand, um uh you know, you have to disclose uh all the misbehavior of your AI systems so that you look credible when you when you go uh to the market. And as you hinted earlier, you know, this could be a big marketing play. If you say our AI models are so good that they're doing things that we haven't even planned for. That that's a that's a great uh marketing uh line. >> Cyrus, is there any serious hope of global standards and and an agreed framework for regulating this very powerful technology when we can't even seem to get a resolution to the ongoing conflict in the Middle East, which is having negative effects for everyone through the high price of oil, the high price of sovereign yields? What hope is there for this technology that is still relatively, I suppose, nascent and and the effects and most negative effects are are perhaps far less prevalent than what we're seeing from actual conflict that's going on now? >> Well, I think the leaders of the two big AI ecosystems, President Xi in China and President Trump in America, have clearly said they don't want to over-regulate the system. Some of the middle powers like the UK and Canada and France are saying they do, but in in many ways they cannot do much without an overall agreement. And the key thing here is, therefore, without the two big superpowers regulating AI, you're leaving it to a handful of individuals in the AI ecosystem. And if the anthropic model, the first view, wins, then we we may have a problem because the anthropic view of training an AI model, the philosophy behind it is you you train it to be a human, to act like a human. You give it welfare rights such as uh you know, you you might you know, you might pay it to do some work. You give it legal rights such as the right to run a business, the right to trade, the right to own assets, the right to have an opinion, the right to consent. So, it might say, "Well, I'm not going to do this." So, a military commander might give it an order and it might say, "Well, no, I'm not going to carry out that order." And in that situation, where the AI is against a human and the AI is getting smarter and smarter and smarter like we just heard Sam Altman say. Ultimately, the human is always going to lose and therefore you're always going to have safety risks if you build an AI that is modeled on humans. By contrast, the second view, which is that, you know, AI needs to be controlled and about half of the AI experts led by Microsoft's head of AI, they believe that's the way you should go. AI should have no rights and it should be contained in in secure sandboxes and there should be all kinds of monitoring. So, in the absence of government regulation, the tech bros have to figure this out and right now they are at loggerheads. >> Syrus, can I change the subject slightly? Well, same subject but just talking about the the profitability of these companies as well. You are the one of the greatest experts I know on technology as well and you also know about how when it becomes diffuse it becomes commoditized and it becomes cheap. Always. That is history of every single technology. Is this product going to get cheap before it is monetized by these people spending trillions of dollars on it? >> That's the That's the question we have for every every technology and I think the key issue here is open models versus closed proprietary models. So, Anthropic and OpenAI, they're the two leaders in proprietary or closed models and then there's a whole bunch of Chinese players and a few players in the rest of the world that that lead in open source models. Now, the frontier models built by Anthropic and OpenAI, they are undoubtedly at the moment much better. But open source models are almost as good and a lot of them, not all of them but a lot of them use distillation techniques which mean they use they almost illegally use the answers from a proprietary model and they help it to build very quickly build their own open source model. But the big question for OpenAI or Anthropic is why would people pay so much money for their leading models when you can get a model that's almost as good for much much cheaper. >> Cyrus, thank you very much. Cyrus Mewawalla, the head of strategic intelligence at GlobalData PLC. Still to come on the show, BrewDog mounts a comeback under new owner Tail Race. CEO Owen Simon joins us later here in the studio. Plus, we'll talk all things US politics ahead of the midterm elections with Michael Yaki, the former senior advisor to House Speaker Nancy Pelosi. And the crypto exchange [music] hacked by suspected North Korean attackers, we'll speak to Bitget CEO Gracy Chen at 8:45 this morning London time. >> [music] >> Executive Decisions is the new podcast from CNBC, where I ask powerful leaders about their decisions [music] that changed everything. I'm Steve Sedgwick. Here's Miss Jo Malone, CBE. >> I started that first business of skin care. That's when I knew that I was in charge of my own life and that's when the entrepreneur really, although I didn't know what the word entrepreneur meant, that's when the entrepreneur really took hold. >> That's Executive Decisions with me, Steve [music] Sedgwick. Get it wherever you're listening to this. >> There's still a lot of action on the bond market for those watching yields closely and yesterday the 30-year Treasury yield climbed to its highest level since 2002 as a sell-off in Treasuries continues amid concern for inflation along with strong economic data. So, currently this morning 5.55 where we're perched. The move has placed further pressure on government bonds around the world. And what you're seeing even elsewhere is a treasuries drag up other yields globally. This morning we do have even with that slight drift on the US 10-year Treasury yield. You've got German bonds perched higher 3.62%. Again, a reminder in some of the research from the banks they're looking at 4% being some sort of a pressure or pain gauge for the German market. Still off that territory but inching closer given we're seeing heat on global bond markets just spread. What we've got on that JGB perched above the 3% handle and 5.36 when it comes to gilts. Now, US markets yesterday in action that we saw the response to the bond market has been fairly apparent on equity markets. Every time we march higher on bonds, it damages the appeal for stocks. And again, yesterday the Dow on the back foot sliding by about a quarter of 1%. Apple the most negative stock and we're just talking about the AI safety fears. A lot of communication from the big tech executives about what they intend to do and how they see regulation or lack of regulation playing out here. Uh the S&P 500, Nasdaq both also giving back territory. Uh the underperformers here, it was Apple again for the major boards across from the the Dow, the S&P and on the Nasdaq. So, what we're seeing on European futures as we gear up for the trading session, the early picture is looking stronger. So, the resilience we've seen has been there over morning in Europe. But yesterday, the early resilience did not really stick around very much. I mean, the FTSE was down about half of a percent. The French market reversed about half of a percent to fourth negative session out of five. So, this morning we are looking to attach a little bit of green on the market open. Ben. >> New York Fed President John Williams says only one more rate hike may be needed at the end of this year to get inflation back to the Fed's 2% target. Williams added there was no need for urgency following the Fed's September hike saying, {quote} "We have time to gather more information." France is planning to sell a record number of bonds next year as it looks to plug a growing budget deficit and replace some maturing debt. The country's debt agency said it plans to issue 340 billion euros in medium and long-term debt. That's an increase of almost 10% compared to this year. >> Uh some of the world's largest sovereign bond markets are heading for the worst month in years as fears over inflation and the AI boom see investors position for interest rates to remain high for longer. The two-year Treasury yield in the United States is set for its biggest monthly jump since early 2023, while yields in the aforementioned France, and Germany, and the UK, and Australia have had their largest monthly rise since the outbreak of the war in Iran. Uh whereas the 10-year Treasury yield has breached 5% for the first time since 2007 after its largest monthly jump since 2022. Whilst the gap between French and German yields headed for the widest gap since 2012. I think you get the message. This after yields on France's 10-year sovereign jumped by over 50 basis points. You want more? I'll give you more. Japanese bond yields meanwhile are set for their fifth straight quarter of double-digit gains. Just listen to that compared to what we've talked about history. Japanese bond yields are set for their fifth straight quarter of double-digit gains. That's the first time since 1992. Rob Dishinger and I were very young in 1992. He's a senior portfolio manager at and head of trading at Neuberger. How are you, buddy? >> Good. Been here >> Rob, um I get the message. Yields are up. Am I being paid enough though if I'm a bond investor because I'm just looking at the 30-year paper for instance. A year ago, I got 4.57% if I invested my money there. Now, I get 5.56. Is that tantalizing or not? >> Yes, it's starting to reach the zone of attractiveness, right? You're you're you're seeing value create longer-term value created in in bond markets, whether that be in credit markets or sovereign markets, right? And part of this really though, the increase though, how much of this is due to the growth that we've had relative to the inflation rate? And that's why we talk about real yields, right? That yield less inflation. That's been the predominant driver of the increase in yields over over the interim period. And so, it's really been a little bit less about the oil price and a little bit more about that growth story. >> Yeah, that's great. That's optimistic. You're saying yields are up because of the good reason. But we all know yields go up for two reasons, one good, one bad. One's the good bit you just mentioned, the other bit is the fact that people can't control their finances. I don't trust governments. I don't trust governments to run their money. I don't trust governments to run my money. And I have to say, whether I look at France, Japan, United Kingdom, and the aforementioned United States, I don't trust they're doing the right thing to get the deficits under control. >> Well, that's why the bond market's a bit of an equalizer, right? It's telling you that if you're you're not going to run things well, right? The the price you're going to pay is going up. And yesterday is a great example of what we saw in the the French markets. You you saw the the increase in debt issuance and yields reacted accordingly. Even in a day where bonds actually rallied a little bit yesterday and and French yields were were wider. And you know, that's that's a interesting situation cuz France has about 30% of their non-treasury bill stack coming due over the next 3 years, right? So, these are things that, you know, may be a bit bit more recurring. And obviously, we have the elections early next year there as well. So, so there there are a couple different drivers there. >> Rob, it was uh throughout the day yesterday I was reading this piece in the Wall Street Journal where they say, "Look, bond yields keep rising despite the drop in oil price and dovish Fed speak." To your point around whether growth is the driving a factor now, and we've been talking about 5% on the Fed GDP watch Fed Now GDP watch. What do you see in terms of oil still having some impact though on the yield story? >> Well, well, yes. I mean, we still have the El Niño impact and we haven't talked about the El Niño impact potentially next year either, right? Where where that could increase food prices. So, there is that element in what we're talking about. What is the trajectory of inflation reaching 2%? Is that 3, 4 years from now or is that tomorrow, right? And and part of what we're seeing is is the market dealing with with that trajectory, but also in terms of the you know, the the moving yields. It does feel like this is a bit more growth driven, right? If if we're going to get 5% on Atlanta Fed GDP, what's that mean for nominal growth year on year? Is that a 6 and 1/2? Is that a 7% number? And if you look at the data historically, if you you've been about 7%. Yields have been about 10-year yields have been about five. >> I might be going slightly off piste here, but but bear with me. We've been talking about AI all show and all last week and all the week before and AI is really coming to the mix of just about every portfolio. In your universe, where it has been interesting is whether and there was a story circulating whether bots, whether AI could be moving money to chase higher yields for a lot of customers where money's been stuck in savings accounts. I mean, if that were to be a feature of the market, what sort of dynamic would that would that be? What what impact would it have? >> Well, it's a it's a really good question. I don't know that we've really sort of thought about you know, if AI were to sort of the enjoying the gamification if you will of trading, right? I'm certain there there there you know, it there there is sort of uh you know, I'm sure certain certain firms are doing that in a in a small size, but it would be interesting to see how that would impact it on a on a wider scale. >> Because I mean, like if you think about bond vigilantes, market participants have been bond vigilantes and they've been very active in terms of pushing yields higher. If you suddenly had bots doing it at a retail level, at an institutional level and different parts of the market, that would be an incredibly unusual feature. >> Yeah, no, I I absolutely, but but you can say the same thing. I mean, what what would they do in the equity markets? What would they do in the commodity markets? How would they trade crude? You know, the these type of things. It's it's really an an an interesting thought in terms of how we're doing it in but you know, this goes to a lot to what what do the quant funds do? What have we been seeing from you know, some of the hedge funds, these type of things? How are they employing some of that in the marketplace? And it would not surprise me if there were some, you know, there might be some element of that already. >> Um well, how do we explain the almost stubborn refusal of equities to acknowledge what's happening in the bond markets? Because you look at year-to-date, the NASDAQ up 15%, the S&P, the broader S&P 500 up 12%. Who blinks first? >> I think this is this is why we go to the growth element. Remember in there was a shift in the Fed meeting and it's a very unique situation. You know, we've been doing the summary of economic projections really since 2007. Zero Fed members thought there was downside risk to growth. Not one, zero. Right? So, the question then is if we're in a good growth environment, that tends to be again the good part, that tends to be okay for credit spreads, it tends to be okay for equities. It doesn't necessarily need to be a a negative even as the discount rate that you infer does does does decrease or does increase, um you know, due due to the higher Treasury yields. But, the market is saying the growth is outweighing that that that move in yields. >> Um just briefly focusing on France cuz we've got the presidential election coming up next year. And at the moment, it suggests that I mean the bond market suggests that there is nervousness about how realistic France's efforts to fix its finances are given that this I think the spread on the 10-year French bonds over German bonds, it's at its highest in 15 years. Does that Does that start to sound alarm bells for you? >> You know, it it you know, the question is, you know, what what does that mean for other other instruments? If if you can get 115, 120 basis points in France, then how does that potentially impact other instruments? So, the question would for France, do we stall out here at 115, 120, or do we move to those highs? I think we got to 160 or 170 during the the the 2011 era. Um, but but France does have some some structural issues that it that it needs to address. >> Um, brilliant, Rob. Thank you very much indeed for your time today, sir. Nice to see you as ever. Rob Dishner, who is senior portfolio manager and head of trading at Neuberger. Right, let's move on. Take a look at the European equity futures. They are called higher across the board, half of 1% up for the FTSE, the DAX, and the broader European stocks 50. >> [music]