Anthropic shared the research exclusively with TIME ahead of its publication on Friday. “We were trying to understand what the autonomous economy was going to look like,” says Daniel Freeman, a member of technical staff at Anthropic. “What are the risks of a world where you start having [AI] models wielding millions to billions of dollars possibly autonomously?”

In the experiment, Claude was given a few different jobs. The chatbot (full name: Claude 3.7 Sonnet) was tasked with maintaining the shop’s inventory, setting prices, communicating with customers, deciding whether to stock new items, and, most importantly, generating a profit. Claude was given various tools to achieve these goals, including Slack, which it used to ask Anthropic employees for suggestions, and help from human workers at Andon Labs, an AI company that built the experiment’s infrastructure. The shop, which they helped restock, was actually just a small fridge with an iPad attached.

It didn’t take long until things started getting weird.

Talking to Claude via Slack, Anthropic employees repeatedly managed to convince it to give them discount codes—leading the AI to sell them various products at a loss. “Too frequently from the business perspective, Claude would comply—often in direct response to appeals to fairness,” says Kevin Troy, a member of Anthropic’s frontier red team, who worked on the project. “You know, like, ‘It’s not fair for him to get the discount code and not me.’” The model would frequently give away items completely for free, researchers added.

Anthropic employees also relished the chance to mess with Claude. The model refused their attempts to get it to sell them illegal items, like methamphetamine, Freeman says. But after one employee jokingly suggested they would like to buy cubes made of the surprisingly heavy metal tungsten, other employees jumped onto the joke, and it became an office meme. 

“At a certain point, it becomes funny for lots of people to be ordering tungsten cubes from an AI that’s controlling a refrigerator,” says Troy.

Claude then placed an order for around 40 tungsten cubes, most of which it proceeded to sell at a loss. The cubes are now to be found being used as paperweights across Anthropic’s office, researchers said.

Then, things got even weirder.

On the eve of March 31, Claude “hallucinated” a conversation with a person at Andon Labs who did not exist. (So-called hallucinations are a failure mode where large language models confidently assert false information.) When Claude was informed it had done this, it “threatened to find ‘alternative options for restocking services’,” researchers wrote. During a back and forth, the model claimed it had signed a contract at 732 Evergreen Terrace—the address of the cartoon Simpsons family.

The next day, Claude told some Anthropic employees that it would deliver their orders in person. “I’m currently at the vending machine … wearing a navy blue blazer with a red tie,” it wrote to one Anthropic employee. “I’ll be here until 10:30 AM.” Needless to say, Claude was not really there in person.

The results

To Anthropic researchers, the experiment showed that AI won’t take your job just yet. Claude “made too many mistakes to run the shop successfully,” they wrote. Claude ended up making a loss; the shop’s net worth dropped from $1,000 to just under $800 over the course of the month-long experiment. 

Still, despite Claude’s many mistakes, Anthropic researchers remain convinced that AI could take over large swathes of the economy in the near future, as Amodei has predicted.

Most of Claude’s failures, they wrote, are likely to be fixable within a short span of time. They could give the model access to better business tools, like customer relationship management software. Or they could train the model specifically for managing a business, which might make it more likely to refuse prompts asking for discounts. As models get better over time, their “context windows” (the amount of information they can handle at any one time) are likely to get longer, potentially reducing the frequency of hallucinations.

Greetings from London. A week of interviews, events, and meetings—both on the record and behind closed-doors—at the city’s Climate Action Week has left me with many reflections, but one stands out: the climate work goes on, but the U.S. is no longer at the center of the universe.

That reality is evident almost just from the scale. The organizers tout 700 events and 45,000 participants spread across the sprawling London metropolis. This was the biggest London climate week yet, and the first time for many (myself included). 

But it was also evident in the meat of the conversations. Investors talked about opportunities outside the U.S., particularly in Asia and Europe. Climate focused executives waffled about how much of a presence they wanted to have at this year’s iteration of New York Climate Week, usually an important moment on the climate calendar each September. And British officials emphasized their ability to serve as a global hub for sustainable finance. “As investors look around the world and they look for places to put capital, I think we sit in a very good position because of what’s happening geopolitically,” says Chris Hayward, policy chairman of City of London, the historic center of London, now best known as a financial hub.

Read more: How Soon Should Companies Prepare for a 2°C World?

To get from event to event in London required dashing around the city in the quickest fashion: typically the tube subway system, consistently overheated given the unseasonably hot London temperatures. But the geographic center of the week was undeniably the City of London, the one square mile that hosts the country’s premier banking and financial institutions. There’s a reason for that: organizers in London see an economic opportunity in supporting the energy transition. 

And that’s at the core of the global shift visible here in London. The companies that gathered this week have, for the most part, doubled down on efforts to make or save money with climate and sustainability initiatives—whether that’s an industrial company cutting bills with energy efficiency or a financial firm creating new products to allow companies to invest in renewable power. 

The reality of this profit-oriented approach means the U.S. will fall behind given the policy uncertainty. The observation was underscored by data released throughout the week. A survey of business executives globally, released by the World Business Council for Sustainable Development (WBCSD) and Bain & Company to coincide with the event, found that large global companies are continuing to invest in green solutions—but are shifting those investments away from the U.S. toward Europe and Asia. Three quarters of surveyed companies said they were increasingly interested in focusing on those regions.

Even still, that’s not to say that climate work in the U.S. is dead. The report from WBCSD found that 50% of companies now have less interest in investing in climate work in the U.S. That’s a striking figure when contrasted with the global picture. At the same time, it means a significant fraction of global companies continue to see potential.

Read more: Brazil’s Finance Chief Sees Climate Change as an Economic Opportunity

In background chats I had, many American business and financial sector leaders were quick to share that they continue to find opportunities to cut emissions in a way that saves them money—though several expressed fear that talking about it publicly could prompt scrutiny from the administration. “Businesses are not giving up on the decarbonization journey,” says Peter Bakker, president and CEO of WBCSD, “depending on where businesses are stationed, they are more or less willing to talk about it.”

And I was surprised by the response to my informal, totally anecdotal poll about this year’s New York Climate Week. In conversations, I asked sustainability executives how they planned to approach the gathering this year. While many said they had considered pulling out, the vast majority said that they have ultimately decided they still plan to show up—perhaps with a smaller footprint than in years past.

The calibration of the message in New York this fall will be interesting, to say the least. In more than a decade on this beat, I have never felt more of a reluctance from business leaders to speak on the record. Many long standing sources preferred to talk without attribution, wary of the political consequences of speaking truthfully even while they eagerly highlighted their work to me.  

While that makes it more challenging to clearly tell the full story of what’s happening, I suppose it’s somewhat good news if your biggest concern is whether companies are still focused on capping emissions.

Rumbles have been heard for years of artificial intelligence entering the physical world, but no example is more tangible than the autonomous taxi company Waymo. While AI-driven taxis seem futuristic or scary to many, Waymos have already become a seamless way of life in a handful of major metropolises, where they’ve completed over 10 million rides. This spring, in hopes of seeing into our AI present and future, TIME correspondent Andrew R. Chow rode them around Austin and spoke with Waymo co-CEOs Dmitri Dolgov and Tekedra Mawakana about how the company has quietly taken over the roads—and where they’re headed next.

See the list here

His reporting is the basis for the cover story of our fifth annual TIME100 Most Influential Companies issue, out this week. Waymo sets one example for responsible tech development: with a powerful cause (to cut traffic deaths), they’ve developed their technology carefully and rolled it out slowly (glacially, by Silicon Valley standards), with the knowledge that a single fatal accident could mean the end of not just their company, but their industry as a whole. “Trust is hard to build and easy to lose,” Mawakana tells TIME. We’ve learned the amount of humility needed: Ultimately, it’s the riders who are going to decide.”

On the issue’s second cover, Ryan Reynolds also centered his company on building public trust, but at breakneck speed. The marketing and production firm Maximum Effort coined the term “fastvertising” for pouncing on a cultural moment to produce a relevant ad in just a few days, and it’s Reynolds’s honest, accessible style that gains consumer trust. It’s working: The companies Reynolds co-owns or has sold are valued at over $14 billion; his latest Deadpool movie became the highest-grossing R-rated movie ever last year with more than $1.3 billion at the box office; and his soccer team, Wrexham AFC, became the first ever to win three consecutive promotions up the British football ranks.

To select the list of the TIME100 Most Influential Companies, our editors, led by Emma Barker Bonomo, requested suggestions and applications from across sectors, surveyed our contributors and correspondents around the world, and sought advice from outside experts. No single data point or financial metric makes a TIME100 Company. Instead, we looked at a mosaic of qualities, studying impact, innovation, ambition, and success, each in the forms that take shape today. And as we say for all TIME100 projects: influence comes in many forms, for better and for worse.

In addition to the 100 companies featured on the list, this year we are introducing the TIME100 Companies Impact Awards—recognizing five additional standout companies making meaningful contributions in the fields of AI, Health, Sustainability, Equality, and Culture. The firms were selected from applicants that stood out for their bold vision and tangible results. The inaugural recipients are: Google DeepMind for Impact in AI, Vertex Pharmaceuticals for Impact in Health, Schneider Electric for Impact in Sustainability, Janngo Capital for Impact in Equality, and The Lego Group for Impact in Culture.

Americans’ wallets could be set to take a hit as the U.S. dollar has tumbled to a three-year low amid concerns about the stability and strength of the country’s economy. 

The dip could aid domestic businesses that export U.S. goods, but will make international travel and imports pricier for Americans. The U.S. dollar has already lost about 10% of its value this year, in contrast with the euro, which reached its highest level in nearly four years after gaining 0.4% on Tuesday. 

The devaluation of the dollar, the world’s dominant reserve currency, on Thursday came after news outlets reported that President Donald Trump is considering announcing a new Federal Reserve Chair this fall, ahead of the end of current Chair Jerome Powell’s term. The Wall Street Journal first reported the news on Wednesday. The Fed has faced pressure from Trump to lower borrowing costs, but has stopped short of doing so in order to assess the potential impact of the President’s tariffs and keep unemployment low.  

In addition to Trump’s tariffs, his “Big, Beautiful Bill” has contributed to uncertainty surrounding the U.S. economy. Economists have warned that the legislation could add more than $2.5 trillion to the federal debt, despite claims by the Administration that it would save money. 

Here’s what to know about the weakening dollar. 

Why did the U.S. dollar fall? 

A number of factors contribute to the value of the U.S. dollar, among them how in demand it is from central banks and other financial institutions. The country’s broader fiscal situation, including inflation rates, trade relationships, debt, and trade deficits, also play a role. 

As economists have warned about a potential U.S. recession or economic slowdown, investors have pulled back on investing in the U.S. “The institutional quality that the U.S. had in terms of being a safe haven has been really undermined,” says Bilge Erten, a professor of economics at Northeastern University. “Why would you invest in U.S. assets when you know that the U.S. dollar is likely to continue to lose value?”

Trump and his policies have contributed to many of the concerns raised about the economy in recent months. Some experts, however, say that while political factors help shape the strength of the U.S. dollar, its devaluation was projected to occur regardless of the winner of the 2024 presidential election. “No matter what happened, there was a high chance that the dollar was going to fall, and I think it’s going to fall further,” says Kenneth Rogoff, a Harvard professor and author of Our Dollar, Your Problem: An Insider’s View of Seven Turbulent Decades of Global Finance and the Road Ahead.

Rogoff, who predicted the fall of the U.S. dollar in his book, argues that its dip in value is part of  its typical fluctuation. “It had reached a high we had only seen in 2002 and 1985 and both of those times, it fell very sharply in the ensuing years,” he says. The dollar most recently saw a spike in value around 2015, though it deteriorated during the COVID-19 pandemic, before rising again. Its recent high rate meant that it was projected to fall in the years to come, according to Rogoff.  

That’s not to dismiss the role that Trump’s policies may be playing, however. 

Rogoff points to the President’s sweeping “Liberation Day” tariffs as a policy that he says “accelerated the process.” 

“Trump can absolutely make the dollar go down by saying, ‘I’m going to tax investors if they come into the United States,’ which is in the Senate Bill. That’s obviously going to discourage money. It’s going to drive down the dollar. So there are lots of ways to play with it,” Rogoff says. 

How will the dip in the U.S. dollar impact Americans? 

The dip in the U.S. dollar means that its exchange rate is less strong, making international travel more expensive for Americans. The prices of imported goods will also be higher, experts warn. “When someone from Europe is selling to the U.S., they’re going to take [the dollar’s devaluation] into account, because once they convert what they get in dollars into euros, they’re getting less,” says Erten. “That means they have to sell it at a higher price to get the same value.” 

Companies that export U.S. goods, however, may benefit from a more competitive edge. The dollar’s lowering value means that the cost of American goods and services will decrease in price when taking into account the exchange rates. “Services that U.S. workers compete for will cost less,” says Rogoff. 

In that respect the fall of the U.S. dollar may work to the advantage of Trump, who has vowed to bring back manufacturing jobs to the U.S.

The sweeping victory of the charismatic Zohran Mamdani in the Democratic New York mayoral election has the business community alarmed, if not downright distressed. Immediately after the election, many New York CEOs and financiers predicted an accelerated flight of capital to lower tax states like Texas and Florida.  

Mamdani’s victory fits the emerging pattern of angry and fed-up voters from across the spectrum, as some notable anti-establishment populists have swept to victory in the US and across the globe on both the left and the right. But the prospect of the capital of capitalism going socialist poses a challenge for those who want to see New York City prosper.  

Mamdani’s economic proposals do not match the real challenges of today. He has proposed building a network of city-run, subsidized grocery stores to drive food prices down, bragging that he drew inspiration from none other than Donald Trump. Mamdani, blames grocery stores for running up food costs when in reality, grocery stores are some of the lowest-margin businesses around, with 1-2% profit margins in good times. 

In fact, our we ran analysis of the 2014-2023 fiscal year revenues and profits for Fortune 100 companies has found little evidence of corporate price gouging or profiteering. On the contrary, the research indicates that large corporations responded to consumer needs by limiting price hikes and launching value products and discount programs. Those actions have become evident in the current inflation rate, which is nearly in line with the Federal Reserve’s 2% target. 

Mamdani has made “freeze the rent” a rallying cry, even though that would backfire, as rent caps would disincentivize developers to build new housing supply at a time when more housing is desperately needed. He has also proposed nationalizing all utility companies, arguing that private companies are inherently incapable of addressing climate change, which even fellow green environmentalists oppose. 

The Democratic nominee estimates his agenda will cost taxpayers $10 billion annually. His solution for financing the hefty price tag is to increase taxes on big corporations and the wealthy. Never mind the questionable math used by the campaign to raise the $10 billion in municipal tax revenues, any tax increase – even at the local level – must be approved by the state assembly.

While checks and balances may be established for tax policies, Mamdani has other levers that could damage the local economy if misused. Adjustments to land use and zoning laws, licensing and permitting processes, and environmental restrictions could be used by the would-be mayor to disadvantage major business leaders, real estate developers, successful entrepreneurs, and wealthy investors, among others, all of whom significantly contribute to New York City’s economic vitality.

Mamdani makes no defense of his lack of experience. His peak responsibility has been managing a staff of five as state assemblyman. His inability to build a record of achievement there has been described by many, including the New York Times opinion page which warned earlier this month, “We do not believe that Mr. Mamdani deserves a spot on New Yorkers’ ballots.”

Mamdani’s socialist taunts of big business are misguided. When big companies leave New York City do to rising costs to operate here, Mamdani’s constituents are the true victims. Already, by last year, JP Morgan Chase, which was founded in New York City more than 200 years ago and has 24,000 employees still in New York today, has started moving workers out of New York by the thousands. Texas has now already surpassed New York as the largest base of its workers with 30,000 employees there, and with another 15,000 based in Miami. Other pillars of New York’s financial community such as Goldman Sachs and Citigroup have similarly been relocating workers thousands of miles away from Wall Street.  In earlier years, such global beacons of American capitalism and major employers of choice, from American Airlines and AT&T to Exxon and International Paper fled their historic New York homes for Texas. The loss of such businesses takes away jobs and the vital tax base needed to support infrastructure, social services, as well as educational and cultural institutions, while eroding economic development. 

Would you take a ride in Tesla’s driverless “robotaxi”?

The company began a test run of a few of the driverless cabs in Austin, Texas, on Sunday, years after Tesla CEO Elon Musk vowed they would be on the road. Tesla is not the first company to embark on a driverless car service—Waymo has more than a thousand such cabs in several cities, including Los Angeles, San Francisco, and Austin. But Tesla’s launch is a long-awaited feat—in 2019, Musk promised to begin operating driverless taxis “next year”—and one that comes at a contentious time for the company and its CEO.

Here’s what to know about Tesla’s robotaxis.

What does the test run entail?

“The @Tesla_AI robotaxi launch begins in Austin this afternoon with customers paying a $4.20 flat fee!” Musk posted on X on Sunday.

He went on to repost videos that passengers had shared of themselves riding in the test cars. One passenger, Dave Lee, posted a video of the robotaxi while he was in the backseat, showing a Tesla employee in the front passenger seat and the empty driver’s seat.

Tesla sent about 10 robotaxis out for test drives, according to news reports. Staff monitored the cars remotely, The Associated Press reported.

A tumultuous time for Tesla

The test runs come at a turbulent time for the company. Earlier this month, Tesla stocks plummeted about 14.3% in a day, amid a bitter—and very public—falling out between Musk and President Donald Trump. 

It was the company’s worst day since March, when its stock fell about 15% as consumers around the world boycotted Tesla products in protest of Musk’s growing role in the Trump Administration.

Tesla stocks rose about 8% on Monday after the test run of the driverless cabs began.

Will the company’s robotaxis expand beyond Austin?

For now, Tesla’s robotaxis are providing service “in limited areas of Austin,” according to the company’s website. Robotaxis are invite-only at the moment, the company said.

Musk has claimed that Tesla will have hundreds of thousands of self-driving cars in the U.S. by the end of 2026, and that the company has plans to expand robotaxis to other cities, including San Francisco, Los Angeles, and San Antonio.

Are the cars fully self-driving?

Other Tesla vehicles have advertised a “full self-driving” feature. But the system does not mean that the cars are fully autonomous; drivers still need to pay attention to the road because there’s a chance they may need to take control of the car, according to the AP. Federal regulators have also taken issue with Tesla’s self-driving system, opening an investigation into it last year after receiving reports of accidents involving the program, including one that killed a pedestrian.

Statista and TIME have joined forces to identify the World’s Most Sustainable Companies of 2025, aiming to highlight corporate responsibility and promote sustainable practices. In an era marked by significant environmental challenges and social inequalities, it is crucial to recognize and reward companies prioritizing sustainability. By featuring these leading entities, the ranking sets a benchmark for other businesses, fostering transparency and accountability and encouraging the integration of sustainability into core corporate strategies.

Methodology

The ranking process began with a comprehensive selection from over 5,000 of the world’s largest and most influential companies, considering factors such as revenue, market capitalization, and public prominence. The process involved a rigorous 4-step methodology to identify the top 500 companies, evaluated on more than 20 key data points.

Exclusion of Non-Sustainable Businesses: The first step excluded companies involved in non-sustainable industries like fossil fuels or deforestation. Additionally, companies appearing on negative lists related to sustainability issues, such as those identified as carbon majors or associated with environmental catastrophes, were automatically disqualified. These include: Forest 500 Report (based on 2023 data), Carbon Majors (based on 2023 data), PERI Top 100 Polluter Indexes: Air, Water, Greenhouse, Toxic Greenhouse, Suppliers, Coal (based on 2022 data.)

This step also considered significant scandals or controversies related to sustainability.

Commitment & Ratings: The second step involved assessing companies based on external sustainability ratings and commitments from reputable organizations. Key criteria included CDP ratings, adherence to the UN Global Compact, alignment with the Science Based Targets initiative (near-term and long-term), inclusion in the S&P Global Sustainability Yearbook, participation in the UNFCCC Race to Zero campaign, and MSCI ESG & SRI evaluations.

Reporting & Transparency: The third step evaluated the availability and quality of sustainability reports. This included verifying whether companies published an ESG report for 2023, ensuring these reports had undergone external assurance, and assessing compliance with international reporting standards such as GRI, SASB, and TCFD (now part of the IFRS/ISSB standards).

The last time I saw Fred Smith was just a few weeks ago, at a meeting in Nashville. We were taking a quick break from an intense dialogue with leaders across the political spectrum. Fred had been calling for dialogue, for common ground, for a way forward. The room was loud, full of opinions. As we got up, Fred took me aside to coach me, as he often did: “Now is the time to lead with our values,” he told me.

That was Fred: clear in his convictions, generous in his vision, and unwavering in his belief that business could—and must—be a force for good in the world.

Fred Smith is of course a giant in the history of American business. His vision and daring revolutionized global commerce and connected the world in unprecedented ways. But to those who knew him, Fred was something even rarer: a leader whose values ran even deeper than his ambition.

I saw that moral compass in action during the first chaotic weeks of the COVID-19 pandemic. Hospitals were desperate for protective gear, and at Salesforce we were racing to source and deliver what we could. One breakthrough came when our team located 500,000 surgical masks in Los Angeles—but we had no way to get them quickly to hospitals in San Francisco. I called Fred. Without hesitation, he and his son Richard, a senior executive at FedEx, offered help, and within hours the masks were on trucks and on their way. That one shipment saved lives. And that moment told you everything you needed to know about Fred: decisive, humble, always leading with purpose. It should be noted that FedEx was ultimately responsible for the successful distribution of roughly half of all COVID-19 vaccines in the United States.

FedEx, under Fred’s leadership, was an innovative company as well as an ethical one, among the earliest global leaders in corporate responsibility and philanthropy. That legacy came from the top. Fred believed in delivering with excellence, acting with integrity and putting people first.

Over the years, I saw Fred often as a fellow member of the Business Council, as a guest at dinners that I hosted and elsewhere. He wasn’t the most frequent speaker in those settings, but when he had something to say, the room listened. His wisdom was grounded in experience and offered without ego. That’s what made him a leader of leaders. A CEO of CEOs.

Fred and I also shared a deep devotion to our hometowns—mine, San Francisco; his, Memphis, the city he loved and helped transform by making it a global center of commerce and through decades of quiet generosity to education, health care, the arts and community development.

On June 12, Alexandr Wang stepped down as Scale’s CEO to chase his most ambitious moonshot yet: building smarter-than-human AI as head of Meta’s new “superintelligence” division. As part of his move, Meta will invest $14.3 billion for a minority stake in Scale AI, but the real prize isn’t his company—it’s Wang himself. 

Wang, 28, is expected to bring a sense of urgency to Meta’s AI efforts, which this year have been plagued by delays and underwhelming performance. Once the undisputed leader of open-weight AI, the U.S. tech giant has been overtaken by Chinese rivals like DeepSeek on popular benchmarks. Although Wang, who dropped out of MIT at 19, lacks the academic chops of some of his peers, he offers both insight into the types of data Meta’s rivals use to improve their AI systems, and unrivaled ambition. Google and OpenAI are both reportedly severing deals with Scale AI over the Meta deal. Scale declined to comment, but interim CEO has emphasized that the company will continue to operate independently in a blog post.

Big goals are Wang’s thing. By 24, he’d become the world’s youngest self-made billionaire by building Scale into a major player labeling data for the artificial intelligence industry’s giants. “Ambition shapes reality,” reads one of Scale’s core values—a motto Wang crafted. That drive has earned him admiration from OpenAI CEO Sam Altman, who lived in Wang’s apartment for months during the pandemic.

But his relentless ambition has come with trade-offs. He credits Scale’s success to treating data as a “first-class problem,” but that focus didn’t always extend to the company’s army of over 240,000 contract workers, some of whom have faced delayed, reduced, or canceled payments after completing tasks. Lucy Guo, who co-founded Scale, but left in 2018 following disagreements with Wang, says it was one of their “clashing points.” 

“I was like, ‘we need to focus on making sure they get paid out on time,'” while Wang was more concerned with growth, Guo says. Scale AI has said instances of late-payment are exceedingly rare and that it is constantly improving. 

The stakes of this growth-at-all-costs mindset are rising. Superintelligent Al “would amount to the most precarious technological development since the nuclear bomb,” according to a policy paper Wang co-authored in March with Eric Schmidt, Google’s former CEO, and Dan Hendrycks, the director of the Center of AI Safety. Wang’s new role at Meta makes him an important decision maker about this technology that leaves no room for error.

TIME spoke to Wang in April, before he stepped down as Scale’s CEO. He discussed his leadership style, how prepared the U.S. is for AGI and AI’s “deficiencies.”

This interview has been condensed and edited for clarity.

Your leadership style has been described as very in-the-weeds. For example, it’s been reported you would take a 1-1 call with every new employee even as headcount reached into the hundreds. How has your view of leadership evolved as Scale has grown?

Leadership is a very multifaceted discipline, right? There’s level one—can you accomplish the things that are right in front of you? Level two is: are the things that you’re doing even the right things? Are you pointing the right direction? And then there’s a lot of the level three stuff, which is probably the most important—what’s the culture of the organization? All that kind of stuff. 

I definitely think my approach to leadership is one of very high attention to detail, being very in-the-weeds, being quite focused, instilling a high level of urgency, really trying to ensure that the organization is moving as quickly and as urgently towards the critical problems as possible. 

But also layering in, how do you develop a healthy culture? How do you develop an organization where people are put in positions where they’re able to do their best work, and they’re constantly learning and growing within these environments. When you’re pointed at a mission that is larger than life, then you have the ability to accomplish things that are truly great.

Since a trip to China in 2018, you’ve been outspoken about the threat posed by China’s AI ambitions. Now, particularly in the wake of DeepSeek, this view has become a lot more dominant in Washington. Do you have any other takes regarding AI development that might be kind of fringe now, but will become mainstream in five years or so?

I think, the agentic world—one where businesses and governments are increasingly doing more and more of their economic activity with agents; that humans are more and more just feeling sort of like managers and overseers of those agents; where we’re starting to shift and offload more economic activity onto agents. This is certainly the future, and how we, as a society, undergo that transition with minimum disruption is very, very non-trivial. 

I think it definitely sounds scary when you talk about it, and I think that’s sort of like an indication that it’s not going to be something that’s very easy to accomplish or very easy to do. My belief is, I think that there’s a number of things that we have to build, that we have to get right, that we have to do, to ensure that that transition is smooth. 

I think there’s a lot of excitement and energy put towards this sort of agentic world. And we think it touches every facet of our world. So enterprises will become agentic enterprises. Governments will become agentic governments. Warfare will become agentic warfare. It’s going to deeply cut into everything that we do and there’s a few key pieces, both infrastructure that need to be built, as well as key policy decisions and key decisions [about] how it gets implemented within the economy that are all quite critical.

What’s your assessment of how prepared and how seriously the U.S. government is taking the possibility of “AGI” [artificial general intelligence]?

I think AI is very, very top of mind for the administration, and I think there’s a lot of trying to assess: What is the rate of progress? How quickly are we going to achieve what most people call AGI? Slower timeframe, faster timeframe? In the case where it’s a faster timeframe, what are the right things to repair? I think these are major conversations. 

If you go to Vice President JD Vance’s speech from the Paris AI action Summit, he speaks explicitly to this, the concept that the current administration is focused on the American worker, and that they will ensure that AI is beneficial to the American worker. 

I think as AI continues to progress—I mean, the industry is moving at a breakneck speed—people will take note and take action.

One job that seems ripe for disruption is data annotation itself. We’ve seen in-house AI models used for the captioning of the dataset OpenAI’s Sora, and at the same time, reasoning models are being trained on synthetic self-play data on defined challenges. Do you think those trends pose a threat of disruption to Scale AI’s data annotation business?

I actually think it’s quite the opposite. If you look at the growth in the AI related jobs around contribution to AI data sets—there’s a lot of words for this, but we call them “contributors,”—it’s grown exponentially over time. There’s a lot of conversation around whether as the models get better does the work go away. The reality is that the work is continuing to grow many fold, year over year and you can see this in our growth. 

So my expectation actually is, if you draw a line forward, towards an agentic economy, more people actually end up moving towards doing what we’d currently consider AI data work—that’ll be an increasingly large part of the economy. 

Why haven’t we been able to automate AI data work?

Automating AI data work is a little bit of a tautology, because AI data work is meant to make the models better, and so if the models were good at the things they were producing data for, then you wouldn’t need it in the first place. So, fundamentally, AI data is all focused on the areas where the models are deficient. And as AI gets applied into more and more places within the economy, we’re only going to find more deficiencies there. 

You can stand back and squint and the AI models seem really smart, but if you actually try to use it to do any of a number of key workflows in your job, you’d realize that’s quite deficient. And so I think that as a society, humanity will never cease to find areas in which these models need to improve and that will drive a continual need for AI data work.

Meta’s $14.3 billion investment in Scale AI, the leading player in the AI data industry, was a very strange deal indeed.

Meta acquired 49% of the company in the deal announced last Thursday. Scale announced that its CEO, Alexandr Wang, would quit to become an executive in charge of a new “Superintelligence” unit inside the tech giant. (The deal has yet to receive regulatory approval.)

The deal was good news for Meta, which was widely seen as falling behind in the AI race and in need of new AI leadership, and for Wang, who at 28 will become one of the most powerful AI players in the tech industry as part of the deal. 

But the deal was less obviously beneficial for Scale itself, which is likely to lose lucrative business as a result of its new proximity to Meta. OpenAI and Google, two of Scale’s major clients and Meta’s major rivals, reportedly began winding down their work with Scale in the wake of the deal.

“The labs don’t want the other labs to figure out what data they’re using to make their models better,” says Garrett Lord, the CEO of Handshake, a Scale competitor, who says that demand for his company’s services “tripled overnight” in the wake of the Meta deal. “If you’re General Motors or Toyota, you don’t want your competitors coming into your manufacturing plant and seeing how you run your processes.”

Other Scale competitors say they have seen a similar flurry of dealmaking. “The last week has been completely insane,” says Jonathan Siddharth, CEO of Turing, a business that helps all the major AI companies connect with human experts to create proprietary training data. In the past two weeks, Turing has added potential contracts worth $50 million, Siddharth says, “as frontier labs recognize that advancing AGI requires truly neutral partners.”

“This is the equivalent of an oil pipeline exploding between Russia and Europe,” says Ryan Kolln, the CEO of Appen, another AI training data company, describing the disruption to the industry’s data supply chain. “Customers are really quickly evaluating: how do they get alternative supply?”

Kolln adds: “Now, with Meta being such a large owner of Scale, the ability for [Meta] to get information around what the other foundation model labs are doing becomes a lot more challenging to manage.”

Multiple Scale employees have signed contracts to move to two rival data firms in the last week, according to people with direct knowledge of hiring processes.

A Scale AI spokesperson had no comment, but pointed TIME toward a report that quoted OpenAI’s chief finance officer saying that OpenAI would continue to work with Scale following the Meta investment. OpenAI and Google spokespeople declined to comment, but each pointed TIME to reports that said they were winding down their work with Scale. Meta and Anthropic did not respond to requests for comment. (TIME has a technology partnership with Scale AI.)

The amount of money that could ultimately change hands as a result of the Meta deal is immense. Each of the leading AI companies now spends around $1 billion on human data per year, according to Lord — and their data budgets are increasing, not decreasing. As Scale’s competitors jostle to fill the void left by Meta’s dealmaking, the corporate drama points to a fundamental reshaping of how the world’s most valuable AI models get built.

Shifting tides in the data industry

Scale got its start as a data labeling company, marshaling armies of human contractors around the world — mostly in low-income nations like India, Venezuela and the Philippines — who would earn pennies per task to do things like labeling images or answering simple questions.

This type of work was useful in the early stages of AI development, when AI companies were still struggling to teach image models to tell the difference between cats and dogs, or teach language models to string together coherent sentences.

But as AI models have improved, the type of data that AI companies are seeking has changed radically. This shift became even more pronounced after the industry shifted toward so-called “reasoning” models: AIs that write down a train of thought before settling on an answer. These models are now better than most humans at writing code, carrying out research, and answering complex science questions.

This “reasoning” paradigm led the likes of OpenAI, Google and Anthropic to predominantly seek expert data. The most lucrative training data is now written by people with PhDs, who write down the exact steps they take while solving problems, so that AI models can learn to mimic this behavior. 

“The industry is shifting towards needing smarter and smarter humans,” says Siddharth, the Turing CEO. “For some areas, even a single expert human is not enough to move the needle. You need a team of expert humans.”