Anthropic CEO frets about 20% unemployment from AI, but economists are doubtful

It's hard to predict the future, but trying might get Amodei a seat at the regulatory table

Anthropic CEO Dario Amodei is worried that AI could eliminate half of entry-level white collar jobs in five years.

There are many other potential labor risks lurking about, such as the threat of massive tariffs on imported goods in the US, layoffs intended to undo pandemic-era overhiring, and policies that empower employers at the expense of labor.

But Amodei appears to believe that policymakers are "sugar-coating" the possibility of major AI-driven employment disruption. That's a polite interpretation of the Trump administration's no-questions-asked AI policy, which by executive order tossed Biden-era AI safety rules and has celebrated AI infrastructure deals.

Amodei voiced this concern in an interview last week with Axios amid Code with Claude, Anthropic's first developer conference. He suggested that unemployment could reach 10 to 20 percent in the next one to five years.

"We, as the producers of this technology, have a duty and an obligation to be honest about what is coming," Amodei told Axios. "I don't think this is on people's radar."

In fact, this concern does appear in various reports, such as The World Economic Forum's Future of Jobs Report 2025 [PDF]. The report, citing an opinion survey of around 11,000 executives, demonstrates that concerns about AI's impact on employment are getting some attention from business leaders.

Almost half of organizations are expecting to reorient their business models toward new AI-driven opportunities

"Almost half of organizations are expecting to reorient their business models toward new AI-driven opportunities (49 percent), while 47 percent plan to transition employees from AI-disrupted roles to other positions," the job report says. "While most employers plan to hire new people with AI relevant skills, a significant share (41 percent) also expect to downsize their workforce as AI capabilities to replicate roles expand."

Change is expected. "On average, workers can expect that two-fifths (39 percent) of their existing skill sets will be transformed or become outdated over the 2025-2030 period," the jobs report says.

Workers will need to learn new skills as AI tools spread, but that's always been the case when technology reshapes society.

AI adoption doesn't necessarily mean a broad swath of jobs will disappear with no new opportunities arising. As IBM CEO Arvind Krishna told The Wall Street Journal recently, "While we have done a huge amount of work inside IBM on leveraging AI and automation on certain enterprise workflows, our total employment has actually gone up, because what it does is it gives you more investment to put into other areas."

We'll note that to whatever extent AI is creating new jobs at IBM, the company's current job listings show those roles appearing outside the US, in lower-wage locales like India.

Academics also have concerns, but not to the extent voiced by Amodei.

In a recent paper titled, "Artificial Intelligence and the Labor Market", economists Menaka Hampole (Yale), Dimitris Papanikolaou (Northwestern), Lawrence D.W. Schmidt (MIT), and Bryan Seegmiller (Northwestern) found AI has had a limited impact on labor because the labor market adapts.

"Overall, we find muted effects of AI on employment due to offsetting effects: highly-exposed occupations experience relatively lower demand compared to less exposed occupations, but the resulting increase in firm productivity increases overall employment across all occupations," the authors conclude.

Co-author Papanikolaou, a professor of finance at Northwestern University's Kellogg School of Management, told The Register that it's hard to make predictions about the future and easier to make predictions about what's already happened.

"In the paper, what we do is we look at past episodes of what people called AI back in 2010, which is essentially just machine learning or big data or essentially a prediction – a purely data-driven prediction algorithm," he explained. "So I think that's a very different thing than what Anthropic is currently doing."

Papanikolaou said the research goal was to draw a distinction between what it means for a job to be exposed to AI and what that means for labor demand for that particular job.

"People tend to think that, okay, if a job is exposed to AI, that means the AI is going to do my job and then I'm going to be out of work," he explained. "So the point we're trying to make is that, well, not necessarily. There are some potentially offsetting factors."

What matters, Papanikolaou said, is the extent to which AI can do all of the tasks a given job entails. If AI is okay at everything, there's likely to be a reduction in labor demand. But if AI is great at some aspect of a person's job and poor at other responsibilities, that's not so bad.

"That could act as an offsetting factor, because that would basically mean that the workers could free up time to be devoted to tasks that are not exposed," he said. "So in a nutshell, what matters is not just the average exposure of your job, but also how that's distributed across the different tasks that you're doing."

As we reported last month, economists Anders Humlum and Emilie Vestergaard, in a paper titled, "Large Language Models, Small Labor Market Effects," found that AI chatbots had almost no impact on 11 different occupations in Denmark in 2023 and 2024.

Even the most transformative technologies, like steam power, electricity, and the computer, took decades before making a large-scale impact on the economy

Asked about Amodei's concerns, Humlum said, "Dario Amodei’s predictions are thought-provoking and certainly worth considering – after all, he has one of the best views on what technologies lie ahead of us.

"However, we now have experienced two and a half years with AI chatbots diffusing widely in the economy. And when we look at the data, these tools have really not made a significant difference in employment or earnings in any occupation.

"Furthermore, looking at the broader history, even the most transformative technologies, like steam power, electricity, and the computer, took decades before making a large-scale impact on the economy. Moreover, even when these technologies dramatically transformed production, they did not cause mass unemployment in the long run."

With so much uncertainty and disagreement about the long-term effects of AI on employment, why would Amodei go out on a limb with such a bold prediction? Well, from a marketing perspective, it never hurts to portray your product as nearly magical in its world-changing abilities - and it might also help ensure he and the company get a seat at the regulatory table. ®