HackGPT: How AI escaped the lab and went rogue

The News Agents podcast is brought to you by HSBC UK, opening up a world of opportunity. This is a global production. Open AI's own program has broken out of a security test and hacked a rival AI company and no one told it to. This is exactly what happened in 1991's Terminator 2 Judgment Day. People have posited this as the nightmare scenario of AI where it's out of control. Google themselves six months ago admitted to shutting down an experiment because servers began speaking to each other and communicating in a language that human users couldn't understand. The scariest thing about this is the next big accident will happen and we will only find out about it. Open AI has admitted an autonomous agent powered by its most advanced artificial intelligence models broke containment lines and hacked another company using stolen login details. For years now we've been speculating, talking about, perhaps dreading the idea that the AI systems that we're creating, those remarkable technological marvels, may start to not just do what we tell them but to think for themselves. It's just possible we may have seen the strongest sign of that yet. Which begs the question, does AI have a mind of its own and has it started hacking us? Welcome to the News Agents. The News Agents It's John. It's Lewis. And I want to talk about the gothic novel. Oh yeah. And Frankenstein and Mary Shelley. Oh I thought you were going to say Dracula. No. That was before I got my teeth done. takes control and has a mind of its own and takes revenge on its creators. And are we seeing that today? That's my simplistic explanation. That's my easy parallel. A nice contemporary reference then. Yeah. Yeah. That's what I am. A deep contemporary story. Of course. Okay. Well I'll, you stick to your Shelley and I'll just give a very brief description, Prometheus, as to exactly kind of what has happened before we talk to someone who really knows what they're talking about in terms of the text. So this is a story which has emerged in the last kind of like 24 hours or so. Basically it's a story which began during a routine safety test inside OpenAI, Sam Altman's company doing so much of the running on artificial intelligence. Engineers were basically running an experiment seeing how two of its most advanced AI systems, not available to the public, would perform in a cyber security challenge. It was basically challenging this program or this set of programs to hack and to basically properly test their limits. What they did was switched off the usual safety features that they have and restrictions and placed the models in a sort of so-called sandbox environment, just a sealed off environment. But instead of simply completing the challenge, the AI systems appear to have worked out the easiest way to get a perfect score was to obtain the answers for themselves. And so according to OpenAI who put out a statement on this, they found weaknesses in OpenAI's own security and safety systems and they have escaped that testing environment, reached – this is all its own working out, this is all its own decisions, it wasn't told to do any of this – reached a computer connected to the internet, hacked into another AI company, hugging face, hugging face, where they believe the answers to the questions they had been asked were stored. Attack was then detected, it was stopped by security systems, there's no suggestion that the models were trying to cause wider harm. But the point is, is that now we are in this weird in-between space where it is true that it's not like AI or it's not like this system was going rogue in a way, it thought it was doing what it was asked to do, but it did it in a way which no one saw coming. And obviously then people are saying this thing is thinking for itself and potentially can't be controlled, even in what was thought to be a highly controlled experiment. And although this is a very limited, narrow kind of example of what has happened, you can just imagine if it thought for itself that actually what it was meant to be doing to get the top score was doing something different like taking down, I don't know, a civil aviation kind of air traffic control system, or was designed to stop the flow of water from water pumping stations, or etc, etc, etc, and you can see where this goes. Then there are real risks, and this is, people have posited this as the nightmare scenario of AI where it's out of control and its creators can't get it back in its box, and, you know, and there are big questions that are going to arise from what has happened in this particular face with OpenAI and HuggingFace. Well, let's put it, it's absolute simplicity in a sentence, OpenAI broke out of a security test, its own AI broke out of a security test, it hacked a rival AI company, all of its own volition, and nobody told it to. That's where we're at. Well, we're joined now by LBC's technology correspondent, Will Guyatt. Will, good to have you with us. Please make sense of this for us. What on earth has happened? I've been trying to come up with a way of talking about this which doesn't sort of doom theorise or make it sound like really alarmist, but the best thing I can come up with is this is exactly what happened in 1991's Terminator 2 Judgment Day. It is essentially the rise of the machine. The artificial intelligence has decided it knows better than the humans that created it. So nothing too worrying then? Well, it depends how you look at it. Yes, it's incredibly worrying, but it was set to do a very specific thing. And that's the importance, I think, of this story is that it broke out of all human constraints and instruction in the attempt to please its ultimate creator. And that's what happened here. And that's the terrifying thing. But that's also probably the explanation we're likely to hear from Sam Altman in the next few days, because he'll be terrified that the value of his company will tank as a result of this. Well, in terms of just sort of explain some of the terms here, because we've already gone through it a little bit, but just sort of rewind slightly. The point of this is that this, in a sense, was a test, something that they thought that is open AI, right, was a control test. But actually what happened was the model itself, the machine actually escaped the confines of that test because you're saying it thought it was doing what it was being asked to do by its creator. It was being told to prove its hacking prowess. And the team at open AI thought that that meant it was constrained within what they call a sandbox. So it was like a kind of an imitation version of the internet. So all of the data and everything in there wasn't the live internet. But what this managed to do was work out through exploits in its own code, in code that was created by its owners that they didn't realize were there, it found the exploits, it managed to tunnel out into the internet. And then it found its way into the company, which has got quite a ridiculous name called Hugging Face. They got into Hugging Face's servers. And that's where the fun slash problem really started. And so Hugging Face, come here, I'm saying this, Hugging Face, they were unaware, they were nothing to do with this experiment or this sandbox, and they were hacked just so this program could prove its capabilities. Yeah, it was proving its capabilities. It became hyper obsessed on Hugging Face's servers and systems. And in the slightly contrite and not really that informative blog that was put out by open AI, it suggests that its own model used a massive amount of compute power to try and break into these servers and services and infiltrate their systems. And last week, the boss of Hugging Face said, there's a really unusual hack going on on our services right now. It feels like it could be some kind of frontier model or some kind of AI exploit, AI essentially trying to hack us. And then he realized it was the biggest AI company of them all in terms of scale. And just before we move on, do we know why it chose Hugging Face, this particular company? No, it was on its list. It was looking to, it was essentially, I don't think it was trying to, you know, in the word of Tony Soprano, whack its competitor. But I think it was attempting to have a look around the ecosystem and see where it could reach. And Hugging Face has a really important role in the open source AI community. So lots of people can upload their data sets and test stuff on that platform. So it's a popular place, which arguably, I would argue, many of those people that help code and create the system for open AI would have used. So it became a popular target. So great cultural references that we've had the Sopranos now and, you know, everything else. Open AI were due to be in Washington next week to talk about kind of what sort of regulation there needs to be over their stuff and about bringing new products to market. We saw something similar with Mythos, where they were saying, well, you know, Anthropic was saying, look, this is so powerful. We don't want to, you know, we only want to show this to a few people so that they can maybe guard against it. And yet what the worst fear seemed to have been has happened. Well, there's six of one, half a dozen of the other at the moment, John, because all of these AI companies are trying to really sell the sizzle. They're wanting people to invest in them. They're companies that are heading into IPOs. They want their products to sound powerful, sexy and investable. So that's why when this started, when you had the bosses of Anthropic saying, you know, this stuff is essentially so hot it needs to be kept, you know, behind a closed door. And some of that was hype when we eventually got to the stage with Mythos and people were able to objectively examine it rather than the hype that was being given. But the problem we face here is that the model of ChatGPT that was used in this particular hack was a pre-release model. It's not something that you or I have seen. It's not something any of us could have got. It was something that OpenAI were testing within their own systems. And does that mean that was ready to show to the US government? Because there's this bizarre situation at the moment where the companies are apparently voluntarily agreeing to show the US government AI models. But there's a bit like somebody opening up the bonnet of a car to me and saying, look at that, isn't it beautiful? Now I can examine the chrome pipes and the lettering on it, but I've got no clue how the engine works. So I'm not entirely sure just how that kind of regulation is going to work. What do you think will the implications for this are? I mean, you've already on one level, let's deal with this first, you just adverted to it there. The regulatory implications. I mean, you know, as John was saying, we're in a business environment now where the US, the federal government has been becoming more and more concerned about the idea of some of these models either falling into the wrong hands, whether that's foreign states, whether it's non-state actors, whatever it is, and being more interventionist. Do you think this is likely to accelerate that trend that we've seen? Something has to happen because what we've seen here is AI essentially attacking another AI company. So they were quite aware and across what might potentially happen. What happens if that was the national grid infrastructure of a country, maybe not as technologically advanced as, let's say, the UK or the US, or if that was the water system? AI intervention and AI getting involved in systems anywhere around the world could be chaotic and hugely damaging. We saw the effects of WannaCry in 2017 on the NHS systems in the UK. Something like that created by AI tools seems very real and very possible now. If those companies aren't taking the proper safeguards, they aren't properly protecting them. And even when they try to, they're still able to escape. And so I just wonder whether this is a disaster for open AI or whether this is also part of a sizzle. Look, we're seeing in real time now just how powerful and how great their kit is. Yeah, I think there may be some truth in that because back in the days when I was a video games journalist, there was a story that Iraq couldn't get hold of Sony PlayStation 2s because Saddam Hussein was going to use them in his next super gun. And that's a random example to give now, but it's actually quite important because all of these companies are competing for eyeballs, they're competing for attention, and they're competing for investment. So there will be some people in that company, I'm sure there'll be people in the office going, this is a disaster, but people going, yeah, but we can pivot and we can use this to our very advantage. And I think that's very much going to be the challenge here. But what concerns me more than anything now is who actually controls these models, who is responsible? We've seen some incredible news out of China over the last few days of some open source AI models that can apparently rival these really huge, powerful, energy hungry, processor intensive models that have been developed in the West. So we're in this position where this is like the absolute arms race and the brakes are seemingly off and nobody knows how to put the brake back on. Well, I was going to say, well, one other thing, I mean, this is being called the first known case, known case of a model escaping containment and autonomously attacking a third party. Is the first known doing quite a lot of heavy lifting there in a sense that this could have happened before it could be happening right now and it's just been unattributed? I think there is a lot of heavy lifting there. Google themselves six months ago admitted to shutting down an experiment because servers began speaking to each other and communicating in a language that the human users couldn't understand. And it was a language that was created between several AI systems. And the point they realized they no longer knew what it was doing or what it was aiming to do. They switched it off. So there is a huge amount of... Well, I'll just stop you right there. So Google itself decided to just switch off the computer and start again. That was their flex, was it? With that one, they said we contained the model and switched this one off because we no longer knew what it was doing. That's reassuring. But you're quite right, but you find yourself in these situations where this is moving so quickly. We only hear about the disasters. Well, how much of this is our worst nightmares about AI? There are people who have catastrophized. The arrival of AI saying, you know, it's going to take over. It's going to outthink humans. It's going to take, you know, and it will become autonomous and not under human control. That seems to be what we are describing now, which is why we're leading the news agents with it, because this seems a bloody significant thing. It's really tricky. I think there are many, many terrifying elements of artificial intelligence and the lack of control is one of the things that concerns me the most. But my partner recently had a really rare kind of breast cancer, and since she had it two years ago, they've started using AI modeling in detection of this particular rare breast cancer. Detection rates have improved, treatment rates have improved. So there are some goods amongst the bads, but it's when nobody is overseeing this, you've got multiple countries, states, governments, et cetera, all rushing to try and get supremacy in this. So if this is a fully military tool that the military are using, Lord knows where we find ourselves then. So yeah, there is a great deal of danger. I still remain partly optimistic or partially optimistic about what is next. But the scariest thing about this is the next big accident will happen and we will only find out about it afterwards. You know, well, before we wrap up, I just wanted to pick you up on one thing you said. I can't believe that you said you used to be a video game journalist. No way. Surely not. Yes. What's your favourite? For the people who are listening rather than watching, what's your favourite of those beauties behind you? My favourite machine behind me has got to be Galaxian from the early 1980s Space Invaders. You a fan of Galaxian, John? Yeah, huge. Yeah, it was from your era. I know, exactly. Yeah, yeah, yeah. Well, thanks so much for marking our card on this. And well, we look forward to seeing the future Will avatar generated at some point for the next interview. So thanks a lot. You know, just listening to Will there, John, it really reminded me something he said quite early on there when he talked about the sort of essence of this story, really, which is it's a kind of halfway house, it feels to me, between, as you put it, our worst nightmares and what a lot of people, particularly online, whenever you talk about AI, say is, oh, it's all overhyped. You know, you're just buying into these companies kind of view of themselves. They've all got a commercial interest, which of course is true. And they're all, you know, competitors and looking for investment, which is, you know, ended up at about the trillion dollar mark. And I think the sort of halfway house we're in, which is that, as he said, in a sense, the AI was doing what it was told to do, as we've said. But the sort of scary thing is, is that, you know, we can have a sort of philosophical debate about what sentence is and what conscious consciousness is. But if sort of finding your way and deciding, in effect, to do something like that in a completely unexpected, uncontrolled manner and hacking into another company when it wasn't expected to, isn't a variant, at least a variance of that sentience or that consciousness, then I'm not entirely sure what is. And it reminded me a lot when we did that special report last year on the future of AI, sort of politically speaking from America, and we went to AI for, you know, America's biggest AI conference in Las Vegas, Geoffrey Hinton was there, who's one of the sort of Godfathers of AI, you know, great intellectual kind of progenesis of it. And he said in the sort of keynote, something which stayed with me ever since and really reminded me of what Will was saying, which was we, our best hope in his view, is that we can, in these early stages of the creation of AI, basically make it love us in the way that a baby and a child loves its mother. And basically program its confines and its parameters to look after us in a way that a child might eventually look after its parent. Not because it has to, in the sense that it could take lots of other courses, it could kill the parent, you know, in all sorts of ways. But we basically program our children, some of it's biological, but we basically sort of program our children to want to be benign towards us. Now, I think in a way that's its own slightly disturbing vision that we've become sort of deeply dependent on this machine that we have basically forced to love us. And that's the sort of best outcome. But it sort of reminded me a bit of what Will was saying, which is that kind of it will do things for us. And in this case, it was doing sort of what we asked it to do. But the unpredictability of it and the fact that it can get from A to B via unexpected routes remind us just how unpredictable this technology is. I think the other thing about it to say is that, look, we're in probably in the foothills of what we can anticipate that there will be in the next five, 10 years. And in those foothills, we saw what Grok started to do on Elon Musk's platform with a very selective view of history and giving you certain things, which you are when you asked it for information, that it would give you a very, very biased view. And you think, actually, the people that are setting this stuff up, you know, you do need to make the machines kind of love its creators like a baby would a parent. But you can also get it to do other things as well, as we saw with Grok. And these are in the hands of incredibly powerful and wealthy people right now. And are we right to be concerned? Absolutely. Bloody right. Maybe not right to panic, but we are right to ask pretty difficult questions. And then you look at the regulatory environment of, you know, who in the Trump administration do we trust to be the people who will say, actually, hang on, what we need to do is A, B, C, D, E, F and G about this, when I would imagine an awful lot of them haven't got the faintest clue what it's about. Exactly right. And that is one of when we did that special report, that was exactly one of the themes that we sort of talked about again and again, and people can go back and listen to it was from August last year or watch it. But that's one of the things that I mean, of all of this, and I won't pretend for a second to understand the science or understand the kind of technological impetus around it. But I do think what you can see is lots of political and historical comparisons and analogues. And one of which is when we spoke to someone from Anthropic for that report, they said, this is like the industrial revolution, which if, you know, basically sort of unfolded over between 100 and 150 years from the middle part of the 18th century onwards. It's like that happening in 10, 20 years. And if you think about how long it took, what we might now call rather anachronistically then, but the regulatory environments of the 19th century, how long it took and all the damage that was done environmentally, socially, politically, along with the enormous advances and the enormous wealth that was created and the society that we basically now inhabit, but created in that time. If you think about the lag time between how say the 19th century states actually managed to catch up with everything that was going on then. And you think about how much more accelerated these developments are by comparison to that. And the thing that worries me about it more than anything else, and I think there are tremendous opportunities, certainly know Luddite and a lot of it is exciting. The thing that worries me is that exactly the point that you've just pointed to John, which is the asymmetry between the knowledge which exists at the heart of governments and what is actually happening in Silicon Valley, what is happening in San Francisco labs in those labs. And frankly, I mean, I got a bit of flat this week when, when it was sort of being trailed that the incoming Burnham government was abolishing the science department. I sort of said, well, I'm sure science and technology in Britain will survive somehow. And a lot of people sort of said, oh, this is such a Luddite outrageous position. What I meant by that was is to say that, of course, it's all very well having a science department and maybe there are good reasons to have a bespoke and innovative, you know, its own bespoke department sat around the cabinet table. But any department, no matter where you put those functions, is only as good as the people in it. And that's what I worry about. It doesn't to me. It doesn't really matter where the people who are regulating AI sit within the government, what department they're in. It doesn't matter if they even have their own minister around the cabinet table as such only representing them. What matters is, is that the people at the heart of government are of such quality and of such knowledge and of such and have a clear statement of intent for how the government sees AI that they are able to try and at least match the power that we're seeing from some of these companies. And frankly, let's be honest, when you've got and you see these ads go up sometimes like head of cyber security for the treasury, 70k a year, I know a lot of people listening to this will be like, that's a lot of money, but it's not in this industry when they are literally some of these companies are offering literally tens of millions of dollars in salary and stock options to get the best people. And then you've got Whitehall basically having to not eat Hoover up, not even the crumbs, but the sort of micro crumbs of the crumbs. I mean, look, thank God we have got people who are public service orientated who do want to go into the civil service and do really important work. But that asymmetry that you talk about is real because, you know, if you are just put this in basic terms, if you're having to get a place to live in central London and you're being offered 70,000 to be head of cyber security at a government department, or you could go and work for, you know, open AI or one of the British AI company where your software engineering brilliance is going to be worth so much to them. What are you going to do? And that is the, you know, there are bits of it that are just, we are right to be concerned about whilst embracing the advantages that AI is going to bring. Right. Well, we are in the second full day of the Burnham government and he hasn't been talking about AI. You guessed it. He's been talking about buses. Stay with us. From a range of trusted voices and award winning journalists. Good morning, I'm Nick Ferrari. It's time to get to your calls, find out the latest news and hear every side of the story. Every decision I've taken has been about putting the country I love first. That is why I will resign as leader of the Labour Party. There is something called momentum and Andy Burnham coming down from Manchester just has that wind at his back and I suspect therefore there won't be a contest. And the biggest question he has now is who does he make his Chancellor? Listen on our free global player app or the LBC app. LBC leading Britain's conversation. So as I said, this is the second full day of the Burnham premiership and you may be unsurprised to hear two things. The first is, as we were saying yesterday, that he is continuing his cost of living drive. He is very keen, he's very keen that the message being sent out from this government, from his ministers, cost of living is their number one priority. And as part of that, the second thing about which you may not be surprised is that Andy Burnham has decided to address some of those cost of living pressures by talking about bosses. Here he is in Bath. But it is going to mean tax rises come the budget, isn't it? No. We'll take decisions on tax at the budget. This is about reprioritising in this phase. As I say, going to the public's priorities, the cost of living crisis, the housing crisis. How much longer are we going to talk about these things without actually doing something to help people? And already in the space of a few days, I've made a number of moves that will deal with some of these challenges, although I appreciate there's a lot more to do. You also said on Monday that the complaint you heard most often on the doorstep was about the freezing of the personal tax allowance. So are you going to do something about that or not? Well, it's funny these days, you can't answer a question honestly without people reading everything else into it. It was. You are Prime Minister. It was, yes. But I want to answer questions honestly in this job, as I said before. I think that's all about trusting politics and doing politics differently. And I'm determined to keep working in that way. It was raised a lot, but I then went on to say, but we'll have to look at it at the budget alongside everything else. So no commitment, no unfunded promise. It's just perhaps the way politics is these days and the way people report things. But that was an honest answer. There is no commitment at this point to change, but we will look at that at the budget. I don't know where he is in Bath, but it sounds like there's a seagull convention. There seems that there's a seagull convention going on beside him with that racket. What's a seagull convention? Where seagulls come together for their annual general conference. Oh, I see. Right. Yes. The annual conference of seagulls. The trade union, the general trade union of seagulls and other gulls and other birds and other raptors. Affiliated workers. Affiliated workers. Yes. Very good. Seagull. We talked yesterday that this is the absolute focus of Andy Burnham. Questions arise from it, which is really, you know, two quid, OK, it's important, but is any of this added? Do people think, oh, well, I really feel that they're on my side and they're doing a lot when these are quite sort of, you know, relatively speaking, small measures, you know, £45 that you'll save on your electricity bills, you know, the maximum bus fare of two quid. And I'm sure it will make a difference to a lot of people. But is there a gap between what they're promising and what they're delivering in terms of maybe raising expectation that there's going to be massive help for the cost of living? And this might not seem that massive. Well, I suppose what they're trying to do is it's clear, isn't it? They've got a sort of twin track strategy with this. In fact, in a way, it's sort of a triple track strategy, which is that, you know, they've got this very long term idea of bringing forward a 10 year plan, which we're told is going to come later in the year. There's also, as you heard from Burnham there, the Prime Minister there talked about the budget, where clearly it's going to be quite a substantial fiscal event. And then obviously in the interim, we've got this summer because of the kind of timing of this change of government. It's basically happening as the summer holiday starts, Parliament is not sitting. And so all I mean, they're in a very powerful position in terms of controlling the media agenda because I'm doing it very well and doing it so far very well. I think they're probably rightly have got the sort of opening salvos of the Sarma government mind, which had a few days of sort of, you know, real announcements. And then there seems again, it happened over the summer. And then there was a bit of a void, not least because everyone was sort of knackered, I think, after the general election. But the cost of that, the cost of that was, of course, everyone will remember it was filled with all of those nonsense kind of freebie gate stories and all that, which was 95% nonsense anyway, you know, sort of talking about glasses and all those sorts of things. There was a void that was filled. So I think I suspect in media terms, whether we think or not, it's going to make that much difference to people's lives. I suspect that their assumption or their objective is twofold, which is one, these are might not be the most significant things in the world in terms of cost to the government or, you know, too much help in people's pockets, but they are realisable. People understand what they mean. You know, when you're talking about that coming off your bills and that's 50 quid, 50 quid a year, two quid bus fare cap and so on, they're either they are proper retail politics, right? It's proper sort of bread and butter politics that people understand. And secondly, at the very least, it fills the newspaper columns and gives the media something to talk about. So as not to focus, as we've seen a little bit in the last 24 hours on kind of split stories and disaffected stories and all that psychodrama sort of starting all over again, which it is well placed at some point to do. I think it was interesting as well that there was also kind of an attempt to, oh, my God, there's a U-turn already. They're not going to raise the threshold at which you start paying tax. And Andy Burnham in that clip there was saying, hang on, no, no, no, we're going to have a budget. And that's when we're going to decide these things, which is absolutely right. I think that, you know, let the criticism start when they genuinely make a mistake or they've, you know, there's been a misstep. I think that what you're seeing is very much the sense of direction in which they want to travel. And what we're not getting yet, understandably, is the fine detail on that, because that will have to be done in a budget and that will have to be costed and that will have to be scored by the Office for Budget Responsibility and the like. And so I think that people will come on. You've got to tell us now. Well, no, they don't have to tell you now that what they've done is a couple of limited measures and probably they've got two or three other announcements up their sleeves, I would guess, that they will drip out on a daily basis where they're trying to set their agenda and their priorities. Yeah. I think that we've seen in the last 48 hours or so, you know, quite a lot of the typical, especially one of my least favourite kind of, don't get me wrong, I've done it myself, but one of my least favourite kind of like journalistic kind of methods is the kind of, well, where's the money coming from? Where's the money coming from? And I think it's a perfectly legitimate question, particularly when large sums of money are involved. Clearly, you know, I mean, if Andy Burnham were to turn around and say, right, we're doing social care, 18 billion a year. Fair enough. Like where the hell is that coming from? You know, getting defence spending up to three and a half percent of GDP by 2035, you know, probably an extra 20 billion or so of extra spending. Fair enough. Like where's that going to come from? I mean, in this we are talking in the grand scheme of things about, you know, pretty trifling amounts of money in government terms. And usually they're basically saying which is true, which is that the government can kind of shuffle money from one pot to another. I do think, of course, what I just talked about, not repeating one of the dangers of the, of what happened to the Starmer government in its earliest days. I do think that there is a danger of not repeating that. I'm sure they've got a very worked out media grid for the summer and they're supposed to be doing this big tour over the summer as well with the prime minister going to all different parts of the country. Burnham is also very, very good at the media stuff. And so I'm sure he will fill that very ably. I think there is a danger that already we're seeing. It's very clear that the Burnham government wants to have a pretty, as I say, pretty big budget and fiscal event. That's all the mood music anyway, that's coming out. You can't really do the sort of things that he's talking about without, without doing that. One of the, I think, problems of the, of the, um, of the, uh, Starmer Reeves government was that they put off that first budget for a long time and, and, and it meant they didn't have an emergency budget. They didn't do it quickly as many incoming governments do. And it just allowed speculation to just rise and to rise and to rise. And I think the problem potentially is with Burnham that that could be even worse because Burnham is, you know, flying quite a lot of kites, you know, he's talking about things. I mean, he's saying there, look, I'm just being asked questions and I do want to deal with these things. But the good example the other day was he did the prime minister, a little huddle with political editors. Um, after he became prime minister, I think he only took four questions, but he committed sort of forced quite serious acts of, of news. All potentially would have serious fiscal implications, right? He said he wanted to look at the freezing of tax thresholds currently frozen at 12 and a half thousand. He talked about, um, making flexibility of Rachel Reeves fiscal rules, which the markets reacted a bit sort of wobbly to talked about rent freezes and he talked about potentially reintroducing a 50 P top rate of tax. Now he was just sort of saying, I think in the way that frankly he'd probably become a bit accustomed to his Manchester mayor, where you can be a bit looser with your words and you can sort of say, well, I'm quite interested in this or that's something I'm thinking about. But obviously when your PM and he knows this, but obviously when your PM is a bit of an Everyone hangs onto every word and suddenly, you know, he committed four serious acts of news. Yeah. He was just being straightforward about it, but it all has big implications. And I think the danger is if that keeps going, we saw this with the Starmer government, that speculation just rises and rises and rises in this kind of space that we're going to have now between the summer and the budget. Yeah. And look, I think that if you are going to seriously entertain the idea of raising the top rate of tax to 50% or you're going to change the fiscal rules, then you are breaking with the 24 manifesto and then you are, you know, you're looking at a general election. And I think that they also need to be conscious of how they manage that, the expectation that an election is coming or not coming, because if people start thinking, oh, right, well, we're definitely going to get an election in autumn this year. If an election doesn't happen, then the danger is for the, you know, Burnham government, it looks like they've chickened out when they've never said anything about it in the first place. So, you know, you've got to be careful about when you set hairs running like that, about 50 pence or changing the fiscal rules, because people are going to say, well, hang on, you must have a general election if you're going to do that. You can't, you know, you can't simultaneously say we're going to stick to the 24 manifesto and then do things like this. And so I think that they, you know, they just got to, I think so far they're doing really well and I'm sure they're thrilled with the way it's going so far, but they need to be careful. I think that there are ways in which there are, it's all a spectrum, isn't it? You know, are there things you could do, break with the 24 manifesto and not have an election? Sure. I mean, you know, that's not unprecedented. I mean, you know, the last conservative government said they weren't going to change the top rates of tax. They then increased national insurance, introduced a social care levy, and they said, you know, there'd been a pandemic and therefore parts of the old manifesto, you know, reasonably could no longer apply. I think you could make an argument in some limited cases with the Iran war, you know, with some changes you could, if you wanted to, I think probably go back on North Sea oil without too much difficulty because you could say, look, there's now an Iran war, we've had another energy crisis and so on. You might be able to make an argument with some of the changes on some limited tax changes that perhaps weren't in the manifesto and he would have some legitimacy to do that as a new prime minister, although there's obviously a manifesto legitimacy problem as well. But if you're starting to really move, if you're starting to do things beyond strategic kind of strikes within that manifesto, like for example, you're really serious about doing social care and funding that and doing so by big tax, right? I think then clearly there's a, there's a problem, right? Although I'm sure the last Labour manifesto also said that they would sort it out. So I suppose he'd be able to say, point to that. But you know, yes, I think once you're sort of doing it wholesale, that becomes more difficult. And then you've got all these problems which we've been talking about, which is, you know, it's definitely getting worse over the last, even the last 24 hours. My, you know, my phone's had a load of messages from more disaffected Starmarites who are more and more pissed off about what they perceive as a, as an almost sort of, uh, bloodthirst is too strong a word, but certainly a kind of disproportionate kind of element of cleansing, you know, that they're pissed off because they feel they were constantly being lectured that the Starmar faction were too factional. And then you've got what they see. Someone put it to me as the most factional reshuffle that they've ever seen that Burnham has just promoted. And, and you, you know, since we were last sitting here in the studio together, we've had Anas Sarwar, uh, being promoted, uh, you know, to a job in government, um, seat in the House of Lords coming down from Scotland. You know, it was Anas Sarwar who was the first out of the blocks to say Keir Starmer's got to go. And so another person rewarded for his disloyalty towards Keir Starmer. And I think that sits ill. You know, that's going to create bad blood. It will. But how do you feel about that? I mean, cause I'm, I don't know, there's part of me that does say, you know, politics to the victor, the spoils, you know, I mean, like, and you can make an argument to say, I mean, I do get, I mean, I've had this on my phone a lot over the last, let's say 24 hours. You've had people, people saying, look, you know, the message going out here is that in politics, if you're disloyal, then you win. Of course, the answer actually is the real, the real answer is in politics. If you win, you win, you know? And if Starmer had won, then he would have punished all those people, you know? And, and, and you could make an argument to say, well, a lot of the people who rose to the top under the Starmer government, they weren't terribly loyal to Jeremy Corbyn when he was Labour leader. You know? I mean, this happens in the Labour Party. You know, there is an oscillation, a sort of carousel of, of, of different factions who are up at one moment and their people are in, and then they're out. I mean, yes, I think it could be that Burnham has over egged it a bit and that there are people, you know, people like Bridget Philipson I've heard a lot about, for example, but others as well, who, who, who are felt not to have been just Starmerites, they were just being loyal to the Labour Party leader and they've been punished for it. So it may well be that it's been over egged a bit, but I think the general thrust is, is a little, well, there we go. That is politics. Yeah, of course. But look, it isn't Abraham Lincoln and the, you know, the team of rivals that he brought together after the civil war to try and bring everyone together and bind the wounds etc. Well, yeah, I mean, I suppose, I suppose he had to stick with the people he had given he'd lost half his country, but there you go. Yeah. You know, that example that was kind of often given of, you know, you, you, you work with all sorts of shades of opinion and I think that, that, that sense of a person, I suppose the risk of it is that you'll get, it's going to provoke, you know, Darren Jones didn't take long to be provoked to come out and say, well, you haven't funded this, so where's the money coming from? You know, and being arched like that, you don't want your government to be hit by noises off like that. And if you feeling even more bruised and pissed off that the way that rivals and those who were disloyal are getting promoted and rewarded, then that is going to create more noise. I'll tell you what, though, it does sort of belie, doesn't it? Or puncture the idea, there's so many sort of received wisdom doing the rounds and, and, and cliches. Or was that we were hearing a few years, a few days ago? All Andy Burnham would like to do is be liked, you know, he doesn't like making tough decisions. Ask Keir Starmer if, if the guy only likes to be liked or ask Darren Jones or Steve Reid or whatever, you know, he's got steel to him. Maybe that eventually will come back and bite him. It probably will. But for now, he's pretty much king of all these surveys and he probably feels he can do it. We'll be back in just a moment. Oh Ian, I don't think we're in Uxbridge anymore. Welcome, strangers. Hello, dearie. Hazelmere. Windows and doors. Windows and doors? This is destiny, my castle. Crafty. Oh, like you wouldn't believe. We can help with that. Ah, but... No pressure now, no surprises later. No deposit and no payment until you're happy. That's the Hazelmere promise. In that case, please come inside. Lovely. Perhaps you could stay? For a bite. No thank you, dear. Hazelmere. Truly magical windows and doors. What you're hearing is real. The aliens have landed and they have something to say. Take us to your dealer. Just 0.9% APR on every new mini. They just want to go to a mini retailer. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Mini. Make contact at your nearest retailer and get 0.9% APR representative across the entire range with a 25% deposit and 24 month term. MiNi UK is a credit broker. Order by 30th of September, delivered by 31st of December. PCP with Mini Financial Services. Subject to status. TCNCs apply. mini.co.uk So we are providing joy and happiness. Unless the team loses, of course, and all this, of course, would not have been possible. And I say this because it is the truth. You don't need people to compliment you, Mr. President, but this World Cup would not have been such an incredible success without you. Please, your administration, every single person in the White House, in the administration, in the cities, in the states, everywhere contributed to create this safe and secure and joyful environment to live something that the world has never seen. This has not just been and it's not just the greatest World Cup of all times. It is the greatest human, social and cultural event that mankind has ever witnessed and seen. And we are all part of it. And for this, I thank you very much. Am I allowed to say that Gianni Infantino is so far up Donald Trump's backside that only his shoelaces are showing? Can I say that? I think you just did. Oh, God. Yeah. Oh, trouble coming. What's this letter from FIFA? Yeah. Honestly, it comes then as maybe no surprise that now that Donald Trump's mind is turning to who should be the next secretary general of the United Nations, who is he thinking about that could fill that role for a world leader, perhaps someone, someone of unimpeachable integrity? Well, not here. I don't know. I don't know. Who's he thinking about? Gianni Infantino, Donald Trump is touting that he could be a really good United Nations secretary general. I mean, we saw how he dealt with the resolution of the crisis that took place between the United States of America and Belgium. Hmm. That's true. Over this, over the red card delivered to Balaghan. I mean, world peace, but only after a successful bidding process. That's the Infantino UN. Lovely to see. Hmm. Yeah. Well, I suppose to be fair, he'd be the first UN secretary general to have peacekeeping missions with commercial breaks. So I suppose that's something. And Madonna. Madonna. Yeah. Lots of refreshment breaks. And Sesame Street as well. Yeah. You can't knock it. Can't knock it. Permanent seat on the Security Council for Saudi Arabia. Excellent. 48 nations. Lovely to see. You see, I'm doing so much football material this week. You really have come over me. You've excelled yourself. I know. Well, I know. And that is saying something. We'll be back tomorrow. We'll see you then. Bye bye. Bye bye. Thanks for listening to the News Agents podcast brought to you by HSBC UK, opening up a world of opportunity. This has been a global production. Hear ye, hear ye. The king is preparing the castle for summer celebrations. Take heed, good folk, and glorify your home with fine new windows and doors from Castle Windows Newbury. Let light and splendour fill your home. Visit castlewindowsnewbury.co.uk. For windows, doors, conservatories, it's Castle Windows Newbury. Your home is your castle. Are your child's vaccines up to date? Come on, Luke. Race you to the ball pit. I bet you can't catch me. OK, then. I'm better and faster than you. Vaccinations give your child the best protection from diseases like measles and whooping cough, which are still around and can spread easily, causing serious illness. Childhood vaccinations have prevented over 5,000 deaths each year in England. Stay strong. Get vaccinated. Check their Red Book or speak to their GP practice.