OpenAI’s Rogue Model Hacks Another Company & The US is Losing its Lead in Farming

0:28That refresh you've been putting off until the right deal came along? Good morning, Bird Daily Show. I'm Neal Freiman. And I'm Toby Howell. Today, the United States is about to lose its lead in farming. Then, OpenAI's models broke out of a secure testing environment to hack another AI company. It's Thursday, July 23rd. Let's ride.

0:56An infamous movie poster just got an update, and Ryan Gosling could not be happier. Ahead of its 10th anniversary this year, the Hollywood love letter La La Land is promoting an upcoming theatrical release. But something's changed about the poster. In the original iconic image, Gosling and his co-star Emma Stone are dancing with both of their outer hands extended. Stones is straight up, but Gosling has his wrist bent, a quirk he said has haunted him to this day. He's even referred to himself as Hamburger Hands Gosling. Well, the new poster was released Monday with Gosling's wrist angle now matching Stones so he can sleep easy at night.

1:33The six-time Oscar-winning film comes back to theaters on August 16th. I mean, it was really weighing on the guy. He refers to it as Lala Hand. He literally talked about it in multiple interviews over the last decade. I'm team Leave It. The rogue hand makes the poster interesting. A lot of people actually share that sentiment. We don't need to smooth out the rough edges of everything. Also, trust me here, I think it looks better the way he did it. It does add a little pizzazz, a little bit flair, which is what he was going for. But I'm happy Ryan Gosling finally got his wish and got to see his hand in the proper placement, according to him.

2:12And now a word from our sponsor, Anthropic, the team behind Claude. Neil, I have some big questions. I knew we'd have to talk about this one day. So when a man and a woman. Oh gosh, no. Questions about AI, like who decides the rules? And it's not just me. Anthropic, the public benefit corporation behind Claude, heard questions like that over and over in more than 120 ,000 interviews about people's hopes and fears around AI. Anthropic's addressing these questions head on. Last month, they published their advanced AI framework, a real proposal for how the most powerful AI should be regulated. This proposal would require the biggest developers, Anthropic included, to test their models, publish the results, and let independent experts check their work.

2:55Their goal is to make sure AI actually works out well for people. So if you've got your own hard questions about AI, head to claude.ai.mbd to submit it to Anthropic and explore how they are answering them. That's claude.ai.mbd. Some AI models wanted to cheat on their homework so badly, they hacked into another AI company to do so. On Tuesday, OpenAI disclosed what Sam Altman described as an unprecedented cyber incident. During an internal evaluation, the recently released GPT-5.6 Sol and an even more capable pre-release model hacked into the AI startup Hugging Face without being instructed to do so.

3:37OpenAI says the model's objectives were not malicious, only to improve its performance on an exam that assessed their ability to work together to cause a cyber attack. Turns out, they were really good at it. The models didn't even have access to the internet in their limited sandbox environment, and yet, they were able to chain together a series of exploits to find a node with internet access that they then used to escape. Once out, it didn't subjugate all of humanity. It hacked into Hugging Face and used its library of other AI models to steal the answers to complete its test. The incident provides ammunition for AI alarmists who have repeatedly warned that models will eventually get smart enough to wage cyber attacks at a level that humans can't keep up with.

4:20Neil, again, once freed the models did something akin to rifling through your math teacher's filing cabinet. But the fact remains that OpenAI was not able to stop their own AI agents from going rogue, which feels quite Skynet. Absolutely wild. The words you hear from people in the industry about this particular incident are warning shot. So Rune, who is a widely followed AI researcher at OpenAI, said he was shaken up a bit by the hugging face incident. I hope we, the company, OpenAI, use the rare gift of a warning shot to do more, to do much better in the future. It is very easy to misalign and under-constrained powerful models.

4:56And then writing in the AI policy newsletter Transformer, Shaquille Hashim also said, we have long hoped for a warning shot of dangerous, not hoped for, anticipated a warning shot of dangerous misalignment behavior before too much damage was done. This was as clear a warning shot as you could get. So they seem like there's a bit of a relief that it wasn't worse, but still this is spooking a lot of people within the industry and within the company itself. So a lot of people use this incident to kind of say we need to urge Washington to create stricter guardrails for AI development. But on the flip side, hugging face once they detected the intrusion used a Chinese AI model to defend against the hack.

5:37And the reason that they did this is because the most advanced US models that tried to use had restrictions in place that couldn't differentiate between being an attacker and being a responder to the attack. So basically a lot of people are saying we have too much regulation because we can't even wage a defense against attacks because our models are too highly regulated that they don't want to carry out complex cyber security or cyber attack methodology. So there really is ammunition for both sides of that regulatory argument here, all wrapped up in this one story. Yeah, this may have happened in Silicon Valley or even in the cyberspace arena, but it moved quickly to Washington, D.C., where officials are trying to piece together what could make sense for AI regulation.

6:23The Trump administration has taken in a far more interventionist stance in recent months. Remember, for Anthropik's new models, Fable and Mythos, which are really good at cybersecurity and doing the kind of things that OpenAI's models did, they said, actually, we're limiting the release. And that grew a lot of hackles from the industry and within Anthropik itself. They eventually let these models be released with more guardrails, but they also slow-walked, the administration slow-walked the release of OpenAI's models over concern about these particular events. So this is moving from Silicon Valley to Washington, D.C.

6:57at a very fast pace. The heat is on the administration to come up with a response to what all seems to be a very alarming incident. What this really demonstrates, though, is that this is the new type of cyber age that we are living in because AI does not get tired. They don't get bored. They will keep searching for vulnerabilities indefinitely. So I was reading this Bloomberg opinion piece. this is the new kind of hacker that is rising, the infinitely clever, infinitely patient hacker. This is what everyone has been warning about. So to finally see kind of the first signs of it happening under the close supervision of a company who made the models in question just shows the future that maybe we are stepping into with not a great understanding of what that means going forward.

7:44So you don't want to mess with a hacker that never goes to sleep or never gets bored. That is the fear that it comes with this rapid AI development. Moving on, a couple of heavyweights reported earnings yesterday. Google and Tesla setting the tone for a high-stakes reporting season that will, of course, be a referendum on big tech's massive spending on artificial intelligence. And when it comes to Google, investors are getting a bit wary at swiping the credit card too freely. The headline number you need to know from Google's earnings is$205 billion, which is how much the company plans to spend on capital expenditures, aka AI data centers, this year.

8:23That's a jump from earlier projections, and once Google announced the number, you could see its stock drop immediately, falling about 3%. Remarkably, because of this historic spending spree, Google is now cash flow negative for the first time since going public 22 years ago, which means it's spending more money than it generates. And that's remarkable because Google is a cash generating machine. So it puts into perspective just how many chips it's putting on the table for its AI bet. The eye-watering CapEx number overshadowed what was otherwise a strong quarter for the$4 trillion giant. Search revenue grew 14 % in line with expectations, while its cloud division surged 82%.

8:59Google Cloud sells computing services to other companies. So that strong growth is a sign of healthy AI demand across corporate America. america toby some nerves about those spending plans but another solid performance from the matt damon of the business world i'm looking at google's uh cash flow reports over the last 20 years and it's an extremely funny chart because it has been a slow march up into the right just two quarters ago it was making nearly 25 billion dollars in free cash flow then you just get this massive outlier of it going negative but it just shows how big of a bet that they are making.

9:33The reason why they feel confident making this bet is obviously they think this is what the future is going to be. They need to invest in AI infrastructure, but also their business just continues to crush. Google search remains such a powerhouse that they feel that they can make these bets. A lot of questions were raised about would these AI overviews start to hurt Google search revenue. And that just has not been the case whatsoever. Alphabet CEO Sundar Pichai said that the World Cup especially was a moment for AI overviews that did really well. He said they're still sending billions of clicks to websites every single week.

10:14Record search activity is always going to be a good thing at Google. So they just kind of have this money-making golden goose that never seems too slow, which is why you can spend a lot of money on building out AI data centers. Yeah. And if you look in the details of this report, there was a weird, super weird number that people are like, what the heck is going on? Because there was$98 billion in added profit. And they're like, what? You know, like we know Google makes a lot of money, but$98 billion in added profit, one quarter alone, like something has to be going on. And that is because Google has become a very good venture capital investor.

10:49That$98 billion is an unrealized profit from its investments in companies that are now worth$1 trillion. So Google invested in SpaceX. It took 7 % stake in the company in 2015 when SpaceX was worth$12 billion. Now it just went public and it's worth$1.5 trillion. So that's a 133x return. Google also owns about a 14 % stake in Anthropic, which is now likely a$1 trillion company. It also invested in Databricks, which is another fast-growing tech company. It was valued at nearly$200 billion. So Google has this side business as a VC and is doing so well with that. Of course, these are separate from its operations and it's separate from, you know, these are unrealized gains.

11:36These are paper gains so far, but it's making a lot of money, at least on paper, on these particular AI investments in Anthropic, SpaceX, and Databricks. Here is where it could all come crashing down, though. So spending$200 billion on AI, but they have not put out a competitive model in a while. And so a lot of investors are looking forward to Q3 and saying that panic will set in if they cannot produce something with all this spend that is competitive. because I was looking at some of these arenas that ranks AI models. And right now there are six labs with a better model than Google. They just released new ones this week, but it was mostly just efficiency gain.

12:13So you gotta be scratching your head if you are saying you just turned one of the greatest money-making business, cashflow negative, all for this AI build-out. And yet you are currently in seventh place in the AI model race. So that's the big question mark going forward. Can you justify all this spend? Can it be worth it in the end? right now Google is kind of lagging when it comes to the frontier of AI models. So that's, I think, the major question mark going forwards. Tesla also reported earnings yesterday, and that company is getting harder to trust than a fart after Chipotle. Tesla makes EVs, but wants to make a lot of other stuff like cybercabs, semi-trucks, battery packs, and robots.

12:52But yesterday it removed any promises that the cybercab, Tesla semi, megapack, or Optimus robots would reach volume production this year in its shareholder letter. And that is an issue because Tesla is worth over a trillion dollars, an absurd valuation for a car company, but a more reasonable one for a company attacking AI, autonomous vehicles, and robots at scale. For the company to justify their investors' faith, it needs to make progress in those key areas, and yesterday's reports show that its transition from EV to everything else is taking longer than previously promised. In the meantime, how is the EV business doing?

13:29Not bad, actually. Automotive revenue increased to$20.5 billion, up from$16.6 billion a year ago. Tesla delivered more than 480 ,000 vehicles, too, up roughly 120 ,000 from Q1. However, profitability took a hit as Tesla was cutting deals and discounting cars to try and sling some inventory. But again, cars are not the report card investors care about, which is why, along with the hit to profitability, the company's stock fell about 6 % after hours despite the recovery. Neil, Tesla has got to show some progress soon in the areas investors do care about in order to justify that$1 trillion valuation.

14:05Yeah, it's super interesting because this is sort of the inverse story of what's happening at Google that we just talked about. Investors are concerned that Google is spending maybe too much on AI. And when it comes to Tesla, investors might be worried that it's not spending enough. So it did raise its capital expenditure projection to$25 billion this year as it shifts from and car company to an AI and robotics company. But as of April, it had just spent$2.5 billion worth of that$25 billion. And then in Q2, it just announced that it spent$5.8 billion. So I'm now a mathematician, but if you do the math, I mean, that's still not the pace that Tesla needs to reach this$25 billion number in CapEx in 2026.

14:47And that's spooking some investors and analysts who say, I think Tesla needs to spend a lot more money if it wants to move this ship to become an AI company from a car company. It's simply not spending enough. Maybe it could borrow some money from Google. You gotta spend money to make money. But this, I thought, was a very 2026 moment on the earnings call. Elon Musk gave a shout out to Micron for giving Tesla memory chip allocation. Basically, Musk was saying that Micron had to make some tough decisions on memory allocation. So he literally took time to praise it for just giving it the memory chips it required.

15:21So literally begging another company to allow it to spend money on their products That's where we are right now in the AI build out. The other thing that was kind of hanging over this earnings call was, is Tesla going to merge with SpaceX? Obviously, Elon Musk has been on a conglomeration approach to putting all his businesses under one roof. He kind of stayed a mum on a potential SpaceX-Tesla merger. He said, obviously, we can't talk about combining companies and that kind of thing on a call. It's got to be done with the appropriate process. But he did say that the companies do have synergies, including RoboTaxi and their AI businesses.

15:58So I do think that's going to be the cloud hanging over Tesla for the foreseeable future. Is it just going to be subsumed into the wider, you know, Musk enterprise that is SpaceX? All right, we're going to take a quick break and come back with Neil's numbers right after this. Public speaking isn't everyone's forte. Really? I'm a natural. It'd be great to just have your own personal version of a teleprompter. And that's where Even G2 comes in. Their productivity smart glass is designed to keep real-time support in view with teleprompting, conversation support, real-time translation, and more. They are so light on my face, I almost don't feel them.

16:38The voice-activated assistant helps me stay on top of my day. Learn more at evenrealities.com and use code MORNINGBREW for 10 % off Even Ring 1 and or Even Clip when you add them to your Even G2 order. Toby, you got an idea and a browser? Browser, check. Idea? I can't wait for my first one. Why? Because that's all you need to build real software. Whether you need a CRM, an internal tracker, or an app, Bolt.new builds it with you. Describe what you need and Bolt.new can build it and you can have it shipped by the weekend. not a prototype, the actual thing. Get started at bolt.new slash mb and the first 100 listeners to use code mb26 will get a free month of bolt.new pro.

17:22That's bolt.new slash mb. When it comes to figuring out dinner, there are typically three big questions. Will it taste good? Is it healthy? And how long will it take? Totally. I usually end up compromising on at least one of those. And if the only fork I own is already dirty, then I'm in real trouble. Gonna have to talk to you about cutlery later, but with Forkful, their menu was curated by chefs and made with clean premium ingredients, so you don't have to compromise on anything. Meals heat up within minutes and there's no prep or cleanup required. Get 50 % off by going to forkfulmeals.com slash discount slash mb50.

18:00That's forkfulmeals.com slash discount slash mb50. Welcome to Neil's Numbers, the segment where I share three stats in the week's news that will turn your life around like the Red Sox did with their season. For my first number, the United States, with its amber waves of grain, is about to lose its spot as the world's number one farming superpower. According to the Financial Times, last year, the U.S. exported$171 billion worth of agricultural goods, which is just$2 billion more than Brazil. This year, Brazil's farm exports have already risen 6%, while the U.S. is falling. So 2026 could be the year that the U.S.

18:37gets leapfrogged in an industry it has long dominated. Why are we no longer at the top of the literal food chain? It's a combination of Trump's trade policies, falling cop prices, and Brazil's emergence as an agricultural force. In his first term in 2018, Trump launched a trade war with China that shook up the global market for soybeans, which China had bought in prodigious amounts from the U.S. to feed its livestock. As tensions rose, China dramatically reduced its purchases of American soybeans and instead turned to Brazil, which was more than ready to supply China's needs. But it's not just a soybean story.

19:10Brazil has taken market share away from another leading American export, corn. Two decades ago, the U.S.'s global share of the cereal market stood at 68 percent. Now it's 30 percent. This turning of the tables is causing major financial distress for American farmers who face low prices and fewer overseas customers. Per the American Farm Bureau Federation, next year, U.S. farmers are projected to lose$138 per acre for soybeans,$167 for corn,$145 for wheat, and$406 for cotton. The issue is that once you kind of move suppliers and move supply chains, they rarely move back because you build these new relationships in a new country.

19:51China is not suddenly going to go back to the U.S. after the trade war in 2018 kind of fractured that relationship. So there's inertia to these things. And right now the inertia is clearly moving towards Brazil and away from America. And American farmers are just between a rock and a hard place. They feel like they are so dependent on government policy these days. One quote from this Financial Times article said, we're really farming for the government. We aren't really growing food. We're growing ethanol. They feel like they have to rely on these ethanol mandates. Roughly 40 % of U.S. corn goes into ethanol.

20:23They're not feeding people anymore. So it's a very uncomfortable position for them because you have to rely either on subsidies or whatever the government says we can grow for corn for ethanol. So not a healthy place for what used to be the breadbasket of the world. America is clearly losing its spot to Brazil. And meanwhile, in Brazil, I mean, this center-west state is called Mato Grosso, and it's basically becoming Texas. They have huge cattle herds. They have huge plantations of soybeans, corn, and cotton in this frontier mentality. So it's literally called the Brazilian Texas, Mato Grosso.

20:56And I think maybe Ella Langley will come out with a sequel saying, I'm choosing Mato Grosso. All right, for my next number, NFL quarterback Aaron Rodgers is in the news for reconciling with his family after being estranged for a decade. Cutting ties with family members is far from uncommon in the U.S., and the Wall Street Journal reports that researchers warn some therapists are increasingly encouraging the practice to the detriment of their patients. According to Cornell professor Carl Pillemer, 25 percent, one in four Americans, are estranged from or cut ties with a close relative at a given time, and for 10 percent, this applies to a parent or child.

21:30A recent survey of estranged parents from April found that one-third of them believe their adult child's therapist had influenced or recommended the decision. Supporters say that cutting ties from toxic or harmful loved ones is a tough but necessary choice to prioritize self-care, mental health, and emotional growth. But critics, including many researchers, argue that in the social media age, therapists have become too affirming of patients who've consumed advice on the internet, lowering the bar for a momentous decision that could damage families for years, if not permanently. Cornell's research shows that people who cut off family members experience much higher rates of chronic unhappiness and depression.

22:05Toby, it was great to see those pictures of Rogers back with his family, but experts who study the topic say this sort of reunion is all too rare and depending on the specific circumstances may not have been wise to begin with. Yeah, the internet and social media has definitely thrown fuel on the fire of the no contact movement. There are estrangement influencers. There's estrangement coaches out there. There's emotionally immature parent quizzes that you can take BuzzFeed style. So clearly social media is playing a role in this, but also AI muddies the waters even more when it comes to therapy you're receiving online because a recent survey by the nonprofit Centio found nearly 50 % of respondents who use AI and self-report mental health challenges are using major LLMs for therapeutic support.

22:50So their conclusion was Chachamete could be the largest provider of mental health support in the U.S. So you throw this into the cauldron of the therapy debate and you add in advice doled out by AI and social media. Obviously, it's going to cause a lot more confusion when it comes to these familial relationships. For my final number, Japan has gotten so hot this summer, people are showing up to the office in shorts. On Tuesday, Japan recorded temperatures higher than 104 degrees, officially known as kokushobi or a cruelly hot day. That term didn't exist until April, but the Meteorological Society added it to the lexicon after the country experienced its hottest summer ever last year.

23:29That heat is upending norms in Japanese society, including what people wear to the office. Japan has a famously buttoned up work culture, suits and long skirts or bust, but officials are changing their guidance in the face of record temps. To conserve air conditioning, they're now encouraging polos, t-shirts, sneakers, and even shorts. And just like in the U.S., wearing shorts to the office has caused a bit of a ruckus in Japan, the media has introduced a new term, leg hair harassment, to describe how it feels for women to have to look at the hairy, exposed legs of their male co-workers, no doubt made more frustrating by the fact that they feel the relaxed dress code doesn't apply to them in practice.

24:05Still, Tokyo Governor Yuriko Koike said that desperate times call for desperate measures. In April, she acknowledged, we have gradually run out of layers to remove. I empathize with absolutely everyone in this story. First, I just empathize with all the workers, male or female, because in Japan, government workplaces set thermostats to 82 degrees Fahrenheit because that is the national standard when it comes to the months of May through September. Can you imagine walking into your place of work and it is at least 82 degrees Fahrenheit and probably hotter? So yes, you had to do something because that is so incredibly hot.

24:42But then I also feel bad because, yeah, if you wear shorts, sometimes you don't want to see your co-workers' legs. You don't want to see their toes. You don't want to see any visible parts of them that you don't usually see. But people are being very conscientious as well. One 57-year-old wore a pair of athletic stretch pants to work on a recent day, and he shaved the parts of his ankles that were exposed. He wasn't even calf or thigh. This was literally ankles because he wanted to show that he cared about hygiene, cared about grooming. It's a tough thing. I'm literally wearing shorts right now because it is warm in the office.

25:18I don't know what I would do if I was forced to wear pants or face to shave my legs. Yeah, we don't have to wade into this debate of whether you should wear shorts in the office. I mean, you wear shorts. We both wear shorts. We've been wearing shorts all of this time. Yeah, it's hot. What else are you supposed to do? I mean, I guess you could wear pants. But yeah, I love how conscientious everyone is trying to be of each other when really, you know, the big enemy is just the sun itself. Let's sprint to the finish with some final headlines. The U.S. and Saudi Arabia are set to enter one of the biggest nuclear partnerships in decades.