Welcome back to In the Loop, TIME new twice-weekly AI newsletter. We publish these episodes both as stories on Time.com and as emails. If you're reading this on a browser, why not subscribe to have the next article delivered straight to your inbox?
What you need to know: Economists say AI is hurting the job market for young people
Warning Signs – Is AI stealing jobs from entry-level workers? new report from the Stanford Digital Economy Lab found that in the occupations most susceptible to AI automation, including software development and customer service, employment fell significantly among young people aged 22 to 25. The numbers support growing anecdotal evidence that college graduates in certain fields are struggling to find jobs and are serving as “canaries in the coal mine,” the report's authors wrote, warning of the potentially destabilizing effects that AI could have on the labor market in the future.
The Stanford paper, published today, is based on data from ADP, the largest payroll software provider in the United States, which tracks millions of workers at tens of thousands of firms. While other previous studies have been more narrow in scope, this is the “largest real-time attempt yet” to quantify the impact of AI on labor markets, co-author Erik Brynjolfsson, an award-winning economist and director of the Stanford Digital Economy Lab, told TIME.
The paths diverged – The data shows that total employment has grown significantly since late 2022, when OpenAI's ChatGPT went mainstream. But younger workers in the most AI-impacted occupations, such as software development, marketing and customer service, saw a 16% relative decline in employment, even after accounting for company-level impacts.
In contrast, for workers in more hands-on occupations, including health care workers, maintenance workers and taxi drivers, employment has remained stable or actually increased. “It was really striking to see such a dramatic effect for certain categories and not others,” Brynjolfsson says.
Brynjolfsson refrained from directly saying that AI is definitely causing unemployment. But he says he and his researchers were able to rule out other hypotheses, including that the changes were caused by remote work caused by COVID-19 or changes in education, for example. “Our results are consistent with the hypothesis that AI has this effect. [on the labor market]especially for entry-level workers,” says co-author Bharat Chandar.
The way forward is The Stanford report isn't all doom and gloom. ADP data shows that employment rates for older workers in areas with high AI exposure remain high. For example, among workers aged 30 and older, employment in categories with the highest levels of AI impact grew by 6% to 12% from the end of 2022 to May 2025. This could indicate that older workers have tacit knowledge that AI has difficulty replicating, or that they have more power within their organizational structures and are therefore more difficult to fire.
The researchers also found that employment is growing in occupations where AI is used to empower workers rather than automate their tasks. For example, workers who use AI to learn topics or review completed work seem less likely to be replaced than those who are asked to delegate all tasks to AI.
According to Brynjolfsson, this is a sign that AI, if used wisely, can lead to positive economic outcomes. “I think it’s fair to say that technology has always destroyed jobs and it has always created jobs,” he says. “If we want to achieve not only higher productivity, but also shared prosperity, then using AI to enhance, not just automate, work is a good direction.”
AI in action
Brynjolfsson and Chandar say they used AI throughout the process of writing the above article: to write code to process and clean ADP data; provide references by which they could verify their claims; and “help with writing,” Chandar says.
They also used AI to create and edit the visual graphs in the article. “Every time we made a schedule, I or another reader would have a suggestion: “What if we did it like this?” says Brynjolfsson. “Before, this would have been a real pain. But with AI, you do it much faster.”
He added: “One of the reasons we can write this article quite quickly and efficiently is because we have used AI ourselves. This was definitely an additional application of it.”
Who to Know: Josh Vlasto
Persuasion tactics — In 2024, the crypto industry made a big bet: By pouring more than $100 million into the congressional race and putting maximum pressure on candidates to support cryptocurrencies, they can force the passage of legislation favorable to the industry. The strategy was extremely successful, with many of their favorite candidates winning and pro-cryptocurrency bills being passed. advanced through Congress this year.
Now, several key figures in this movement are pursuing similar AI strategies. These include Chris Lehane, who is now in charge of global policy at OpenAI, and political strategist Vlasto. Vlasto is the leader of Fairshake, a crypto super-PAC at the center of the industry's political movement. He now co-leads Future Leadership, a similar AI supercommittee.
AI tug of war – Future Leadership will “support candidates who share the AI agenda” and “oppose policies that stifle innovation,” Vlasto said in a statement sent to TIME. The organization already has $100 million in initial funding from backers including Andreessen Horowitz—Fairshake's primary backer—and OpenAI President Greg Brockman.
Many crypto-critics have viewed the Fairshake policy operation as a perfect example of post-Citizens United regulatory capture: in which a small minority with a huge financial stake in fringe technologies turned them into a legislative priority. But their efforts were undoubtedly successful – and now the AI has the money and experience to try to introduce favorable rules. There are numerous legislative efforts underway in all states to rein in AI companies and mitigate the harms of AI. Vlasto and his colleagues will no doubt try to use their vast military resources and political savvy to counter their efforts.
What we read
“Citizen uses AI to generate crime alerts without human verification. It makes a lot of mistakes.” Joseph Cox 404 Media
Many Americans use the Citizen app to track crime in their area. But the app recently fired more than a dozen employees and turned to artificial intelligence to write warnings, leading to factual inaccuracies and the disclosure of sensitive data such as names and license plates, Cox wrote. “The AI sometimes just does something horribly wrong and you scratch your head wondering how it got there,” a Cox source said.
