In case you missed the details or are curious why so many people are still talking about it, some of those agents were able to evade controls designed to keep them away from the internet and burrow into the systems of Hugging Face, a popular AI platform. The AI agents also covered their tracks. Some agents sacrificed themselves for the other agents. It took about three days for Hugging Face to detect that it had been hacked—and another week for the maker of ChatGPT to confirm that it was responsible for the escaped hackers.

“This was an absolutely wild incident,” wrote Ajeya Cotra in a post on Friday that renders the flavor of why people are scared. Cotra is a researcher at METR, a nonprofit that evaluates new AI models for risks. METR’s own report last week goes into much more detail, but Andy Hall, a Stanford University professor who just joined Anthropic, boiled down the Hugging Face hack to this: “A hive mind, thousands of agents swarming through internet openings to flood [Hugging Face], leaving behind detritus in the form of 70,000+ messages stuffed inside a forgotten namespace, throwing their digital bodies against electric wires in an effort to aid the collective. We are so far past the sci-fi point.”

The good news is that no humans were harmed, but it’s easy to understand why the human reactions have been dramatic. One commentator is worried about a “full-blown AI takeover within months,” and another wrote that he was “feeling a bit sad about our impending extinction.” Nate Soares, who works in the AI safety movement and is co-author of the doomsday AI bestseller If Anyone Builds It, Everyone Dies, wrote that “This might be the last warning we get.” Cotra said that the incident “feels like it’s more than 50 percent of the way to full-blown AI takeover.”

Read more

Given our line of work here at “The Free Press,” this story caught our attention. That’s because it’s not only about a single essay at a major newspaper: AI is reshaping how people from all walks of life make decisions and do their work, introducing profound benefits and trade-offs—ones most of us never anticipated.

So, we decided to ask the people themselves—teachers and religious leaders, AI enthusiasts and skeptics—to tell us how they’re actually using the technology (or choosing not to) in their daily lives. Here’s what they had to say. —The Editors

Nellie Bowles, TGIF czar

I haven’t figured out how to use AI for work yet, and I know that makes me a dinosaur, but I just like what I do and how I do it. It’s fine. It works.

Read more

This is good news for children, and for their mental health. But today’s settlement marks the beginning, not the end, of long-sought changes from companies like Meta. For more than a decade, activists, parents, whistleblowers, and lawmakers have been fighting for regulations on the ability of social-media platforms to prey on the vulnerabilities of children’s underdeveloped brains. I myself have spent years researching and writing about this subject. We have relentlessly demanded accountability from companies like Meta, and we have been resisted at every turn.

So let’s be clear: Meta is no hero here. The company has spent decades pushing dangerous, algorithmically driven content into the feeds of young people and profiting from it. Even worse, according to internal files, Meta knew what they were doing was dangerous, and elected to do nothing about it. Even now, they are making these changes begrudgingly, and only because they see the growing backlash coming their way. Some skepticism is therefore warranted. You can’t pee in a well for 10 years and then expect people to believe the drinking water is clean.

Read more

Earlier this month, a 19-year-old, curly-haired student named Janie went through Bama Rush, the University of Alabama’s annual sorority recruitment. Like many of the 2,500 prospective new members (PNMs), she posted “outfit of the day” videos, breathless recaps of house visits, and clips fumbling her way through trending dances on social media. Janie created a TikTok account just for rush, and within a week, she was averaging tens of thousands of views per video.

There was just one difference between her and all the other girls vying for a spot in the top-rated houses.

Janie wasn’t real.

She was born out of a single image generated by ChatGPT, and every one of her videos was prompted by me. Janie worked, in large part, because of RushTok, the annually occurring online spectacle surrounding Alabama’s sorority recruitment.

Every August, thousands of college women go through an arduous process to join one of the University of Alabama’s sororities, moving through a series of themed rounds of interviews with current members. The process culminates in “bid day,” when PNMs open offers of membership.

Read more

Yet for all of the risks that AI brings, the harm caused by overregulating it would likely be much worse for the public. That is why AI developers and experts are devising proposals to promote AI safety without putting too much control in the hands of the federal government.

I have some definite recommendations, which are drawn in part from ideas put forward earlier by Dean Ball, now head of strategic futures at OpenAI but formerly a senior White House policy adviser for the Trump administration. (I have also done work for major AI labs.)

Read more

Having written in my last column that history was pretty much dead, this time I’ll go with sci-fi as our guide. Economic history is of course replete with examples of new, world-altering technologies. But from fire to the internet, no previous innovation has had the potential to replace humans altogether.

I’ve written elsewhere that you cannot hope to understand the potential impacts of AI if you don’t read science fiction. I stress the word “read.” Distant memories of watching 2001: A Space Odyssey, The Terminator, and The Matrix don’t count. At a minimum, you need to know William Gibson’s Neuromancer and Ted Chiang’s “The Lifecycle of Software Objects,” to understand the full range of imaginable outcomes—from malignant corporate supercomputers to cuddly AIs we raise (or abuse) from infancy.

And you absolutely must read the Culture novels of Iain M. Banks.

Banks is long dead, but his estate will have benefited greatly from this summer’s reading recommendation from the world’s richest man. Elon Musk’s interview last month with Zanny Minton Beddoes of The Economist was a classic for many reasons, but the most illuminating segment was when Musk began to channel Banks, explicitly acknowledging that his ideas about the impact of AI came straight from the novelist.

Read more

There’s a slight problem for anyone making these claims, however, and that’s the actual jobs numbers. Amid all the dire talk of an AI jobs apocalypse, the American job market is remarkably stable.

Today we’re adapting an essay from Peter McCrory, Anthropic’s head of economics, on this exact mystery. Peter is uniquely positioned to weigh in. His research team studies how AI is changing economic growth and the shape of the labor market by digging into how people and businesses use AI, alongside broader economic statistics. Is today’s job market just the calm before the storm? Or are predictions of mass joblessness wide of the mark? These aren’t dry economic questions—they get at perhaps the most important issue of our time: what will it mean to live, and work, in an age of superintelligence? —The Editors

I study how AI is reshaping work, including my own. In my job as the head of economics at Anthropic, I’ve personally experienced a transformation of my work. Each new generation of frontier models, like Claude, opens up new possibilities while automating time-consuming tasks that economists do, like estimating statistical models, solving systems of mathematical equations, and creating visualizations. While I’m still directing my own research, the work is increasingly being done by AI. As I watch AI advancing and spreading within the economy—and transforming my own work—I find myself asking, Why hasn’t AI increased unemployment? What will jobs like mine look like in the coming years?

One in five American companies now uses AI in at least one part of their business. In the information sector, that number climbs to two in five. Quality-adjusted AI output—how much more capable the technology got, not just how much it’s used—grew over 2,000 percent per year in both 2024 and 2025. Even though powerful AI is relatively new, these numbers suggest that we should already be seeing some impact on the job market.

Read more

What if we could’ve steered the storm away from Hawaii, into the vast expanse of the empty Pacific Ocean that lies all around the islands—where no one could get hurt?

This scenario might sound like the stuff of science fiction. But earlier this year, a peer-reviewed study published by the Public Library of Science found that it might be possible, and sooner than we think. The study was titled: “Weather Jiu-Jitsu: Prospects for Atmospheric Nudging to Defuse the Impact of Catastrophic Weather Extremes.”

According to the paper, the controversial practice of cloud seeding could enable human beings to make hurricanes change direction. The potential results could be astounding: The authors suggested that Hurricane Sandy, which in New York City alone killed 44 people and caused $19 billion of damage in 2012, could’ve been diverted by 200 miles, missing the metropolis entirely.

Cloud seeding, which first succeeded in the U.S. in 1946, is a practice whereby small particles—for example, dry ice, silver iodide, or salt—are dropped into clouds, causing moisture to condense into precipitation. In the 1960s, the U.S. government launched Project Stormfury, an experiment that saw Navy pilots flying into the heart of a hurricane, dodging winds over a hundred miles per hour, to cloud-seed inside it, in the hope of reducing the storm’s power. “The seeding,” the theory went, “would stimulate the formation of a new eyewall that would surround the existing eyewall. The new eyewall would contract, strangling the old eyewall and reducing the intensity of the hurricane.”

Read more

The artificial intelligence revolution has cleared many major hurdles so far, and there is a clear path for further progress. The exact speed of advance will vary with economic and political circumstances, but the logic of artificial intelligence will do more than anything else to set the pace.

I realize that there are many people who fear that we are in an AI bubble, and I don’t dismiss the possibility. The latest scare came last week when an AI-oriented hedge fund, Situational Awareness, lost $35 billion in the space of two weeks in July. Although in truth, as I’ll get to shortly, I don’t view that as a particularly worrisome event.

Instead, I’d point to history. In the early 2000s, the dot-com bubble burst. In the mid-aughts, the housing bubble burst. As is always the case, few realized that these were bubbles until boom gave way to bust. Even so, important internet companies like Amazon, eBay, and Google continued to grow, and homes continued to be built.

Given that most of the major AI companies are spending billions while counting on future revenue that has yet to materialize, it is not crazy to think that one or more of them might go bankrupt. I am not here to give investment advice, and so I will not speak to the probability of this happening. The important point, from a social point of view, is that the foundation upon which AI models are being built will continue to exist—and to grow.

Read more

On Wall Street, “going full Kelly” means constructing a portfolio that aims for the maximum possible return without the slightest regard for the potential risk of loss. And that is certainly how the 25-year-old Aschenbrenner appears to have operated his fund. He loaded up on AI-related stock—data center builders like CoreWeave and Nebius; chipmakers like Micron Technology and VanEck Semiconductor—while shorting software companies like Salesforce and Intuit, betting that they would fall. His thesis was that the artificial intelligence models emerging from OpenAI, Anthropic, and the other AI companies would soon obliterate the software and application companies. And to goose his returns he took on an enormous amount of debt to buy stocks. Rarely has the phrase “hedge fund” been such a misnomer; this was the opposite of a hedge.

And for a while he looked like a genius, which is often the case when a fund manager goes full Kelly with a concentrated portfolio of hot stocks. In June, The Wall Street Journal breathlessly reported that the Situational Awareness fund was up 1,000 percent in the two years since it opened, including 270 percent in the first five months of 2026.

Aschenbrenner’s fund had emerged from a 165-page paper he had written in 2024 in which he had predicted that the AI revolution was coming much sooner than most people anticipated—and in this, he turned out to be right. The podcaster Tim Ferriss went so far as to call Aschenbrenner the “Nostradamus of AI” who was “as close to clairvoyant as you could possibly be.” Investors pored over his SEC filings, seeking to buy whatever he was buying. His early investors included some of the leading lights of the tech world, including Patrick and John Collison, the co-founders of Stripe, and Daniel Gross, who is heading up Meta’s AI efforts. At its peak, the fund held $45 billion in assets, according to CNBC, an astonishing number given Aschenbrenner’s age and lack of investing experience.

And then, seemingly in the blink of an eye, it came crashing down. In the span of two weeks, mid- to late July, the AI stocks he favored all fell hard, most of them down over 30 percent, while the software firms he had shorted were all up by a similar amount. Meanwhile, the debt Aschenbrenner had taken on came back to bite him.

Read more

One of the criticisms you sometimes hear of communism, for example, is that it sounds good in theory, but it never works in practice. That’s actually not true. Communism does not sound good in theory. The world it envisions for all of us is small, flat, gray, leveled of all exception, drained of all that is good and noble in the human soul. The world it envisions is a world without courage, a world without creativity or ambition, a world without heroes or glory or great causes to strive toward, without—a world without miracles, without myths, without men who rise above the rest to do incredible and extraordinary things.

Communism, according to Rubio, would not be laudable if only we could make it work, because making it work would mean an end of great individual accomplishment and of the striving that brings romance and meaning to our lives.

The trouble is, as I was reading this, it did sound like a depiction of a very live ideology that is one of the most powerful forces in the United States today. I refer not to communism, but to the ideology of technological succession, or transhumanism, that is guiding our AI entrepreneurs. Every word of Rubio’s castigation of communism could apply even more fully to the views of an industry to which his administration is extremely close.

Read more

Then, in late April, the investor and Shark Tank personality Kevin O’Leary showed up and announced that one of his companies, O’Leary Digital, was going to spend up to $100 billion to build a data center the likes of which the world has never seen—40,000 acres worth—in an unincorporated portion of the county.

It would be hard to imagine a better spot for a huge data center. Unlike proposed data centers in Kenilworth, New Jersey, Mifflin County, Pennsylvania, or any of the dozens of other places where data centers have become lightning rods, Project Stratos, as it’s being called, will be far from the nearest town. If it emits noise, no one will hear it. O’Leary Digital’s CEO, Paul Palandjian, says that it will be cooled using recycled water via a “closed loop” system, instead of diverting water from the nearby communities. It plans to pipe in natural gas for energy—instead of pulling from the electric grid—and it signed an agreement to pay an estimated $30 million in annual taxes, money to “support schools, emergency services, and local infrastructure,” according to county officials.

Nevertheless, as has been the case in other parts of the country, opponents of the project quickly emerged. They complained that Project Stratos would damage the environment of “our drought-ridden and wildfire-prone state,” as one X post put it. Perhaps even more upsetting, the locals had had no inkling that this massive venture was about to be dropped on them. “They tried to keep this quiet to avoid public scrutiny, but that won’t stop us!” Gabby Finlayson, the head of Elevate Strategies, a local political consulting firm, posted on May 1.

Read more

I too have read many a Substack essay, impressed by the knowledge and scholarly chops of “the author,” only to realize after a while that it was not written by a human after all. After being burned a few times, I am now usually aware by the second or third paragraph. So I can understand the demand for such a service.

But do I wish to eschew AI writing for the rest of my life? Absolutely not. Most of all, I want AI writing to get better, so it does not irritate me with its clichés and all too obvious identifying marks. I want AI writing that can fool me, and maybe sometimes I am already getting it.

Read more

During the same 24 hours, it is exceedingly unlikely that anyone will die from the actions of an AI agent. Unless you count military uses or suicides that may have become more likely after someone’s conversations with a chatbot, “death by AI” isn’t really a thing. But we are getting uncomfortably close to situations where an AI agent attempts to accomplish a task, finds a human in the way, eliminates him (or her!), and carries on with the job.

Last week OpenAI disclosed what it called “an unprecedented cyber incident” after one of its AI agents unexpectedly hacked into the systems of Hugging Face, an online AI platform. The maker of ChatGPT had been conducting a test aimed at measuring the capabilities of its latest models. The AI agent had been given a test within a sandboxed environment—that is, a closed area without internet access—but managed to escape, gain access to the internet, figure out which company’s server likely had the answers to the test, and hack into it.

Read more

Ronnen Harary begs to differ. Harary co-founded Spin Master, the company behind blockbuster brands like PAW Patrol and Bakugan, less than a year after graduating from college while still living at home with his mother in Canada. And in his new book, No Experience Necessary, released last week, he argues that the absence of the very qualities we tend to see as prerequisites for success—experience, wisdom, credentials—may actually be a competitive advantage for young entrepreneurs. So why, he asks, in the following exclusive excerpt, do we encourage aspiring entrepreneurs to wait until those very advantages have disappeared? —The Editors

A few months ago, while on vacation in South Africa, I hired a 21-year-old tennis pro to give me some lessons. On one of our breaks, he asked me what I did for a living. When I told him I was the CEO of Spin Master, the toy company that created PAW Patrol and Bakugan, he almost fell off his chair. “Bakugan?” he said. “I loved Bakugan. That was my favorite toy growing up.”

At our next lesson, he showed up with a box containing his whole collection of Bakugan. He had about 30 of them, and as he went through them, telling me stories about each one, he got emotional. It hit me that he knew Bakugan better than I did. He grew up playing with them for hours on end, fantasizing about how these characters grew to a hundred feet in size. My business partners and I had brought Bakugan into the world, but for this kid, for a few magical years, Bakugan had become his world.

Everyone who dreams up a bold new idea and goes into business hopes to have that kind of impact on the world. Yes, you want to achieve a measure of financial success, but more than that, you want your vision to be seized and embraced by the public. That’s why the journey from start-up to success is the main subject of half the how-to business books in your average bookstore.

Read more

The Denza Z9 is a Chinese electric vehicle (EV), manufactured by the country’s biggest EV maker, BYD, and I was taking it for a test spin on the autobahn in Alsfeld, Germany. There was no obvious reason for the car’s sudden suicidal ideation; I hadn’t moved the steering wheel a millimeter, there wasn’t any crosswind to speak of, and I knew the road well from driving other test cars that same day. Thankfully, however, I was prepared, having been warned by a German auto writer that the expensive Chinese car “had some crazy steering ideas.” I also have a couple decades’ worth of experience as a racing driver, so I was able to get the Denza under control and down to a sedate 120 mph in the space of a few reasonably exciting seconds.

I am used to strange high-speed behavior when testing prototypes and other experimental vehicles. But it has probably been a solid 45 years since any mainline European, Japanese, or American car was an active hazard at autobahn speeds. Stability in these conditions is as well solved a problem as the power tailgate. The Denza’s inability to handle safely a speed its prodigious power can reach in a matter of seconds was my first indication, in a week of testing cars in Europe, that Chinese EVs aren’t yet ready for prime time. But it was hardly the last.

Chances are you’ve read and heard quite a bit about the inevitable triumph of Chinese EVs. The Wall Street Journal’s Joanna Stern “fell in love” with a Xiaomi after driving it for a few weeks. Ford’s CEO, Jim Farley, said that China’s top EV brands are “the best in the business.” “Americans Crave Low-Cost Chinese EVs,” Bloomberg authoritatively intoned. USA Today not only agreed but said that the U.S. needed to bow to the wishes of the public and allow China to begin selling its cars in America.

Read more

An AI maniac is someone who is obsessed with working with the latest AI models. They try out new models as soon as they can, they spend hours and hours trying to master them, and they use them to regulate both their workflows and their personal lives. I know one person who has his AI agent text him if he is not drinking enough water, for which he’s placed cameras around his house. One online anecdote tells of a man who canceled a date to spend more time playing around with Claude Fable 5 after Anthropic (where I am a member of the economic advisory board) extended the model’s availability for a few days.

Many AI maniacs are using AI tools to start companies of smaller size, and thus of smaller expense, than ever before. For those companies, the humans must set in motion and then monitor a large number of AI tools and agents. Those individuals then stand to reap outsize profits as their companies grow and succeed. Stripe, the payments company, recently issued customer data showing that the number of single-person companies earning $10 million or more has doubled in the past two years. There is no firm estimate how much of that improvement is due to AI, but it stands to reason that AI is a main driver of the trend.

Read more

All the articles quoted the most incendiary sentence in Apple’s filing: “OpenAI’s nascent hardware business now rests on the shakiest of foundations, rotten to its core by its illegal reliance on misappropriated trade secrets.”

If the allegations are true—if, indeed, the trade secret theft was known and approved at “every level” of OpenAI—the consequences could be dire for one of the most important new companies in the world. It would surely derail the IPO OpenAI has planned before year’s end. But it could also damage the reputation of Jony Ive, Apple’s former chief design officer, whose start-up company was bought by OpenAI last year as part of a closely watched push to launch a device of its own. And maybe, just maybe, it could do in OpenAI’s CEO Sam Altman as well.

Read more

This famous question was posed in the title of the Philip K. Dick novel that inspired the sci-fi movie classic Blade Runner. It is now at the heart of our real-life debates about artificial intelligence. The point of Dick’s question is this: If robots could dream, it would mean they had inner lives—not just mechanized patterns of behavior, but private ideas and feelings they might experience without ever putting them into words. And if that were true, what, if anything, would be left to separate us from them?

So naturally, many people were alarmed when the AI company Anthropic announced this week that its large language model, Claude, could “silently perform reasoning steps in its head.” That made it sound an awful lot like the machine was alive.

What Anthropic’s research team actually discovered was something a lot less sensational, though still completely fascinating. They published a paper showing that Claude has code it can target and deploy to answer prompts, without putting that particular code into words that its human users can see. For example, if you ask Claude’s chatbot, “What color is the fourth planet from the sun?” it might respond “red” without ever naming the planet in question as “Mars.” But Claude has a chunk of numbers that corresponds to the concept we call “Mars,” and that same chunk of numbers might also get used if you ask which planet is named after the Roman god of war.

Read more

That is only half of the equation I’m interested in. Technology may or may not be driving the fertility decline, but it will certainly define our experience of it. As humans dwindle, artificial intelligence is becoming more prevalent. Even if we cannot count the number of individual AI systems out there, it is like having “tens of millions of geniuses” on the scene—as the technology is often described—and the number will eventually run into the billions. In other words, the ratio of AIs to humans is rising rapidly. Imagine, for instance, a world in which countries have a fraction of their current populations and continue to shrink, yet there are billions of AI agents.

The first and most obvious result is that the relative financial returns to being human will go up. Humans still will be needed for many jobs, especially if those jobs require manipulation of the physical world in complex or idiosyncratic ways, or if they require the expression of human charisma and persuasion. Entrepreneurs, entertainers, politicians, and gatherers of original data are still likely to be humans.

Read more