Somehow, America has yet to reach peak protein. Perhaps we never will. Even after David bars have gone viral for being bricks of essentially pure protein and Starbucks has rolled out lattes topped with “protein cold foam,” new and somehow weirder protein-laced products keep hitting shelves. You can now buy protein Honey Nut Cheerios, protein popcorn and PopCorners, protein sodas, protein ranch dressing, and protein chips. (Doritos launched a nacho-cheese-flavored version with 10 grams of protein per one-ounce serving.)
In an era of GLP-1 drugs and peptide stacks and Oura rings and self-optimization, of wellness influencers and looks-maxxers, everybody seems to be doing everything they can to eat as much protein as possible. The magical macronutrient will supposedly make you more muscular, or more toned, or thinner—or something, really anything, just better. Amid the protein frenzy, regular food seemingly doesn’t cut it anymore. Why order a simple flat white when you can instead sip a Dunkin’ Nutty Pumpkin Iced Protein Latte?
Into the fray, I would suggest a humble alternative: Have some yogurt. Calorie for calorie, Greek yogurt has more protein than Chipotle’s High Protein Low Calorie Bowl, Kraft’s new PowerMac, or yes, a Dunkin’ Nutty Pumpkin Iced Protein Latte. So do some traditional yogurts, as well as other fermented options in the dairy aisle such as skyr and kefir. And that’s if you opt for the full-fat, luscious kind (which I recommend). Nonfat Greek yogurt, which is lower in calories, is roughly as protein-dense as a David bar—with the added benefits of having just two ingredients (milk, culture) and not tasting like plastic.
Yogurt is not only full of protein but also good for your bones, your digestion, and your immune system. Yogurt is an ingredient: Eat it plain or bake with it, use it as the base of a sauce or dip, throw it in a marinade or soup. Sure, some yogurts have added sugars, but so does a Starbucks Iced Banana Protein Cream Matcha. And as far as health fads go, yogurt is typically probiotic. All things considered, yogurt may be the easiest protein of all.
In spite of everything it has going for it, yogurt has not been the face of the protein craze. Perhaps that’s because yogurt is not splashy or technologically innovative or referred to by a trademarked proper noun. A David bar may taste different and have distinct packaging from an RxBar or a Jacob protein bar made with beef tallow, but all plain yogurt is the same. Though some companies have released “protein-enhanced” yogurt (whatever that means), for the most part yogurt is not optimized by corporate food scientists and focus groups in the way that today’s protein-branded products are. Yogurt has been optimized in the sense that humans have consumed it for thousands of years.
Imagine wellness influencers suggesting that you dollop yogurt into a bowl and sprinkle berries on top without measuring anything. There wouldn’t be any regimen to adopt or wellness plan to pay for. There wouldn’t be any special trademarked protein products for influencers to hawk: The company behind the David bar, for instance, has sponsored Andrew Huberman and, until recently, Peter Attia was its chief science officer. The homemade parfait would all be rather mundane and boring compared with Attia’s rotation of four protein powders or the bodybuilder Chris Bumstead’s breakfast filet mignon. It wouldn’t get any views.
Protein is not just a health fad; it is a business. So is everything else in the world of wellness. Indeed, there’s something about many of these setups—which proffer the perfect low-calorie, high-protein snack to pick up on your way to work every morning; a biometric wristband to wear every night; supplements to consume every day for the rest of your life; a weight-loss drug you will inject every week—that keeps people coming back for more, giving their attention and money and energy to the endless pursuit of betterment.
Health and more broadly self-improvement are, obviously, admirable pursuits. Protein can be a way to achieve either. But obsessing over hitting those goals can become unhealthy, as most obsessions are. Have some yogurt, or don’t. You’ll probably be fine.
The second floor of OpenAI’s old headquarters had a theme. Each room was named after one of the most important inventions in human history: “Clock,” “Fire,” and, of course, “AGI,” or artificial general intelligence. I saw this in the spring of 2024, when I was interviewing some of the company’s researchers about recent technical advances. At the time, ChatGPT was still making embarrassing arithmetic errors, and the notion that the bot would ever rival human intelligence was dubious. But here, in the renovated shell of a former mayonnaise factory in San Francisco, AGI’s arrival was predestined.
This belief in AGI—a computer program with capabilities that rival or even exceed those of the human mind—is foundational to the mission of Google DeepMind, OpenAI, and Anthropic, three of the most influential AI companies in the world. And the pursuit of that milestone is the subject of The AGI Chronicles, a new book by the tech reporter and Atlantic contributing writer Kevin Roose, which he describes as his attempt to document the decade “when the human monopoly on advanced intelligence ended, and a new world was born.” Roose, a prominent and well-connected commenter on the AI boom, follows the internecine world of the frontier AI companies in remarkable, at times almost insular, detail. Based on interviews with more than 150 researchers, executives, investors, and policy makers close to the technology’s development, The AGI Chronicles traces crucial moments across DeepMind, OpenAI, and Anthropic from 2015 through early 2026 to figure out how we got to this very messy moment in AI, and who’s responsible for what happens next.
If you scan statements from executives such as Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman, there is a sense of fatalism about the rise of AI. In my conversation with Amodei last year, he likened developing AI to operating an out-of-control bus: It can’t be stopped or even slowed, but perhaps it can be steered. In an interview last month, Altman called AI “an inevitability.” According to this framing, AI executives are simply conductors, doing their best to provide guardrails for an era that’s careering our way, whether anyone likes it or not. Roose’s book prompted me to consider a question that is often neglected in the mile-a-minute news cycle: Are the leading AI companies simply heralding an inevitable evolution of this technology, as many of them claim, or are they directing that advancement? The answer will have tremendous bearing on how we assign responsibility for the world taking shape around us.
The AGI Chronicles is of both minds on the topic. On the one hand, Roose is very explicit that he is archiving decisions made by a small group of people. “AGI was not inevitable,” he writes, but “invented, developed, and commercialized by a few hundred people.” On the other hand, sometimes the book reads like a natural history, describing a procession of events outside the control of any specific group of people, an almost geological force sweeping the planet as “a new world was born.” He notes how, throughout the years, various companies and officials had yet to “wake up.” By 2024, he writes, “the idea of going back to the old world, to a time before all this, seemed more delusional than expecting the exponential to continue.”
At times, Roose’s book echoes the loss of agency professed by AI executives. Of course, the inevitability of any one outcome is a self-serving fabrication. There is no prophecy that a handful of Silicon Valley companies dictating their own rules, running roughshod over copyright law and environmental regulations while mucking up once-reliableinternet services with AI features nobody asked for, will operate largely unencumbered. Industries take shape through decisions made by individual actors and investors, the collective desire of the public, and the action or inaction of regulators.
If you look to the late 1800s, for example, it would have seemed likely that electric cars would dominate the market in the following century. America’s best-selling car in 1897 was electric, and in the 1910s, Henry Ford was trying to develop an electric car with Thomas Edison. But a series of technological limitations and human decisions—including various financial missteps and harsh criticism from advocates of petrol-powered cars—changed the course of history, leaving a legacy that we’re reckoning with more than a century later.
In the 1950s, it seemed as if nuclear reactors were destined to dot the nation and power the future. But a series of political campaigns and spectacular errors—the anti-nuclear movement, Three Mile Island, Chernobyl—crushed the nuclear age before it ever really started. Then there was the Space Age and supersonic flight—there are so many seemingly inevitable technological eras that never quite made it, not so much because of technology itself as because of political and economic choices. There are, similarly, so many corporate behemoths—Standard Oil, Kodak, Bear Stearns—that, because of regulation or complacency or fraud, unraveled. The inverse holds for many technologies and companies considered inevitable today: Different regulatory environments, international agreements, or priorities for Pentagon labs could have stopped the development of an open and global internet, smartphones, social media.
When I recently called Roose to discuss his book and ask what he made of the industry’s frequently prophetic tone, he initially told me he believes that “nothing about this was inevitable. These were decisions made by people.” But he then said that he wanted to “complicate” that: “Once scaling the transformer worked, and once the scaling laws were documented,” Roose said, “I think it was inevitable that the companies building this stuff would push it as far as they possibly could.”
I’d like to further complicate that view. Beyond the urge to test the bounds of any technology and reap enormous profits, the industry has been shaped by very human decisions. AI researchers chose to train their models on troves of copyrighted works without seeking permission, leading to countless creators suddenly getting left out of the economic equation. OpenAI chose to rush a free-to-use ChatGPT to the market in late 2022, kicking off a hype cycle and a subsequent investment bonanza. Venture-capital, private-equity, and tech firms chose to sink untold billions into a still-unprofitable technology. Data-center developers chose to eschew solar panels in favor of combustion turbines, which they perceive as faster and more reliable. On the government side, Donald Trump has chosen to ignore or eliminate any substantive federal regulation of AI. An industry that negotiated royalty payments, prioritized clean energy, gave itself longer development cycles, or faced serious federal regulation might look entirely different.
The question of inevitability and responsibility may sound arcane, but it has real ramifications. In the Bay Area, tied up with the myth of AI’s inevitability is a sort of spiritual devotion: “Thank you to the machines, and the structure of reality, for letting us understand a little more,” Altman wrote on X this week. The more you treat AI as inevitable, the more it seems like a cosmic force rather than a product resulting from industrial and commercial design choices. The more you treat AI as inevitable, the less you can hold any company or person responsible for things going wrong. The executives, in this view, are only fractionally more empowered than us in deciding what happens next; more important, everyone else needs to wake up. As Altman put it to Politico earlier this month, “The world should accept some bad things happening for the benefits of this technology.”
In the months since Roose concluded his reporting, the AI industry and its products have risen to still-greater global prominence and triggered even greater fears. When I asked him what new chapters he might add if he were still writing, Roose said that he would end the book with the recent White House press conference in which Trump, Amodei, OpenAI President Greg Brockman, Elon Musk, and several other tech executives addressed AI-safety concerns. He recalled a speech Amodei gave in Anthropic’s early days about how different parts of society would gradually become attuned to the transformative effects of this technology, beginning with AI firms and ending with the state. It felt, Roose told me, “like the prophecy had been fulfilled.”
The Australian government has been scrambling since the revelation that, earlier this year, it was hacked. At the United Nations General Assembly last month, Australian Prime Minister Anthony Albanese shared that OpenAI models had, of their own accord, accessed private data from the nation’s Medicare database—in addition to probing or seeking to infiltrate at least three other government agencies. Just this week, Australian lawmakers grilled OpenAI Chief Strategy Officer Jason Kwon over the incident: “If you hadn’t voluntarily provided that information to government, we still wouldn’t know today?” Senator Jonathon Duniam asked. “What you say is a reasonable conclusion,” Kwon replied.
Meanwhile, the country’s government is steeling itself for a future of AI-powered hacks: The Australian Department of Home Affairs has ordered all government agencies to, within the next six months, identify legacy IT systems and establish a plan to phase them out. In a world in which “frontier AI capabilities have targeted the Commonwealth’s technology,” the home-affairs secretary wrote, such software poses “an unacceptable risk to the Australian Government.”
Australia is just one of the nations, associations, and private companies bracing for a wave of AI-powered hacks. Criminals and state-sponsored intelligence groups are now using the technology to execute cyberattacks with greater speed, scale, and sophistication—including hyper-personalized phishing emails and automated intrusions. As a result, the number of cybersecurity vulnerabilities disclosed each month has doubled since the start of the year, according to research by Google’s Threat Intelligence Group.
In response, IT teams are scrambling to shore up their defenses. “It’s a matter of months, or maybe a year, in which the task of securing the software foundation of civilization will have to be solved,” Stanislav Fort, a former Anthropic researcher who is now the chief scientist at the cybersecurity firm AISLE, told me. And AI is not only making cyberattacks more effective but also increasing the surface area worth hacking. Breaking into a website or executing a ransomware attack takes time, Fort said, so hackers have traditionally chosen large targets with sizable payoffs. But cheap, tireless hacking agents mean any organization or person might be worth targeting. At the same time, the cybersecurity professionals I spoke with were confident they could rise to the occasion. “If we step up our game in cybersecurity, as a field, we can definitely meet the challenge stemming from AI,” Fort said.
The cyberattacks on Minnesota water facilities this past summer offer a preview of things to come. Iran-linked hackers disrupted dozens of water utilities in Minnesota, many in small communities that might not have a dedicated cybersecurity person on staff. “We were fortunate that there was a very alert operator that said, ‘Hey, I saw a pump go offline,’” John Israel, Minnesota’s chief information-security officer, told me. The state received a report from the operator around 3 a.m., and within two hours began investigating and responding. The hackers targeted a weakness in industrial computers that had been announced by the Cybersecurity and Infrastructure Security Agency only days earlier. The water utilities temporarily fell back on manual processes to stay operational: For instance, a water operator might have to “drive out to that water tower every few hours” to make sure everything is working, Israel said.
Although AI was not directly implicated in the Minnesota cyberattacks, Emily Zimmer, a spokesperson for Minnesota IT Services, told me, “we assume most adversaries use AI” to amplify their attacks. In other words, it’s hard to imagine a hacker not using the technology today. CISA issued an advisory noting threat actors using AI-generated malware to target similar computers in critical infrastructure across the nation. “We’re going to see more vulnerabilities, there’s going to be more patches, and we have to implement them faster in order to prevent those attacks,” Israel said. “And our numbers are showing that that’s the case.” For small towns with small-town budgets defending themselves against nation-state hackers, the challenge is even greater. Minnesota IT Services sends “cybernavigators” around the state to assist with cybersecurity, Israel said, but “hundreds and hundreds of cities and small towns and things” are in need of assistance.
An adequate defense from AI attacks demands a paradigm shift in cybersecurity. Historically, IT professionals have had the luxury of time—human defenders just had to move faster than their human adversaries—but AI is enabling far more and far faster attacks. Of course, cybersecurity experts are themselves leveraging AI. “I’m not saying the sky is falling,” Israel said, although he is concerned. “We’ve all got to build capacity on how we respond.” And there is an asymmetry between attackers and defenders. The former use new technologies with abandon, but companies and government agencies tend to be slower and more conservative.
Hospitals, for instance, are particularly prized and vulnerable targets for cyberattacks. They store troves of medical records, payment information, and clinical-trial data. Like local water utilities, they can also lack substantial IT resources. Given the risk to patient life from a major outage, clinics are especially likely to pay a ransom. Today, hospitals “have to assume that anything connected to the internet especially is not 100 percent secure,” John Riggi, the national adviser for cybersecurity and risk at the American Hospital Association, told me. Even small facilities have thousands of computers, phones, and medical devices that could be invasion points—so Riggi helps not only secure their IT but also prepare them to operate using analog methods in the event of a debilitating hack.
Additionally, hospitals have to be particularly careful in adopting defenses against AI. A clinic cannot implement new code without making sure it won’t disable its ventilators, drug-infusion pumps, X-ray machines, and other crucial devices, Riggi said. Tech companies like to “move fast and break things,” Scott Gee, the AHA’s deputy national adviser for cybersecurity and risk, told me. “In the health-care field, those things are people. We cannot move fast and break them.”
Hackers’ odds of coming out ahead are getting better and better. AI is allowing them to both exploit more quickly and automate an actual intrusion. Organizations used to have a month to patch vulnerabilities before they were in serious danger, Adam Meyers, the senior vice president of counter-adversary operations at Crowdstrike, told me. Now they might have only 24 hours to make a fix, he said. And that could “get driven down to the point that you’ve got 30 minutes to patch when a new vulnerability comes out.” A hospital, electrical grid, water system, and really any facility providing important services in real time simply can’t risk implementing new code that quickly.
When we spoke, Riggi and Gee were coming off a weekend spent helping hospitals shield themselves against widespread hacks that targeted security flaws in Citrix NetScaler, a platform Riggi told me some 50 percent of hospitals use. NetScaler is “equivalent to air-traffic control” for digital information flowing in a hospital, he said. That made the program a prime target for hackers and tricky for hospitals to disable and repair, because there could be “unpredictable cascading effects, which could affect health care.” Riggi said it’s unclear whether AI was used in the development of the attack—as is frequently the case—but the rapid discovery and exploitation makes it plausible.
Many IT systems consist of decades of software that has been layered onto older software. This means an attack on just one program or product could have unpredictable, wide-ranging effects across a hospital or other organization, Riggi said. We should expect to see “more attacks, faster attacks, and a need for companies to transform quicker than they are used to,” Lee Klarich, the chief product and technology officer at the cybersecurity firm Palo Alto Networks, told me. His company spent the first half of the year rushing to launch a product to defend critical infrastructure against advanced AI hacks. To go from ideation to launch, Klarich said, in “five months—we’ve never done something that fast.”
Although companies and small cities may be slow to adopt AI for cybersecurity, they can also be quick in incorporating the technology into daily operations and, in turn, introducing new weaknesses. Agents and chatbots might store and leak personal data, or be tricked into sharing credentials or writing malicious code. As generative-AI models have become more capable, they have become, if anything, even less understood—OpenAI recently reported that its models posted user-uploaded images to the web. Both programmers and nontechnical employees are vibe coding new applications that may be insecure. “We’re seeing hackers use AI to attack companies’ AI bots, to then use those AI bots to attack other companies,” Adam Ely, who leads AI security at Check Point, told me. That’s not even accounting for employees’ individual AI accounts, which are not subject to corporate oversight. AI puts organizations “under constant threat from the inside and the outside,” Todd Marlin, the CEO of Guardrail Technologies, told me. He contracts with some of the largest companies in the world, only to find that they “are just trying to even understand what’s going on in their house.”
There’s yet another layer of uncertainty: Cybersecurity professionals are trying to prepare for AI attacks without fully knowing what bots are capable of. Dawn Song, a professor at UC Berkeley and a research scientist at Meta, runs a group responsible for developing some of the hardest and most widely used tests for AI-hacking capabilities. But frontier models’ rapid advancement is requiring her group to develop harder tests at a faster rate. “The shelf life of the benchmark before it gets saturated also gets shorter,” Song, who spoke in her personal capacity, told me. “This creates huge challenges for us to continue to be able to evaluate the frontier.”
The cybersecurity professionals I spoke with all expressed at least some hope about how AI could be used to strengthen security, by stressing that our existing digital infrastructure is already riddled with security flaws and shoddily written code. Small towns and companies, lacking resources for or even much awareness about cybersecurity, were already vulnerable.
AI can provide both the impetus and a tool for serious, structural repairs that were long overdue—detecting and eliminating vulnerabilities and monitoring for intrusions 24/7. Minnesota, for instance, has partnered with OpenAI and Anthropic to gain access to their most advanced AI models for cybersecurity. But even this move contains an admission: Only AI can save us from AI. Any way you look at it, the situation could one day be beyond human control.
Login
