Dario Amodei, CEO of Anthropic, is warning that AI could reshape the labor market much faster than many workers, executives, and lawmakers expect. His central concern is not only that artificial intelligence will change jobs, but that it could remove a large share of entry-level white-collar work before society has prepared for the shock.
In comments to Axios, Amodei said AI could wipe out half of all entry-level white-collar jobs and push unemployment to 10–20 percent in as little as one to five years. He pointed in particular to tech, finance, law, and consulting as sectors where early-career roles may be exposed.
A warning about entry-level white-collar work
Amodei’s forecast focuses on the bottom rungs of professional office work. These jobs often involve drafting, analysis, research, coordination, document handling, and other tasks that language models are increasingly being used to support or automate.
The risk, as he frames it, is that companies may adopt AI quickly enough to reduce hiring for roles that once trained new workers. If fewer entry-level positions exist, the effect would not be limited to people currently holding those jobs. It could also change how future workers enter fields such as tech, finance, law, and consulting.
Amodei argues that the public conversation has not caught up with the speed he expects. He told Axios, "Most of them are unaware that this is about to happen," adding, "It sounds crazy, and people just don't believe it."
His most striking scenario combines broad progress with severe labor disruption: "Cancer is cured, the economy grows at 10% a year, the budget is balanced — and 20% of people don't have jobs." The point is that technological and economic gains, in his view, do not automatically protect workers from displacement.
Why Anthropic’s CEO says AI leaders should speak plainly
Amodei leads Anthropic, the company behind advanced models like Claude 4. He says that position creates a responsibility to talk openly about where the technology may be heading, even when the message is uncomfortable.
He told Axios, "We, as the producers of this technology, have a duty and an obligation to be honest about what is coming." He also said, "I don't think this is on people's radar."
That warning sits inside what he described as "a very strange set of dynamics." According to Amodei, AI companies can warn that their own technology may create serious labor market consequences, while critics may respond that those companies are exaggerating the threat for attention. His reply to skeptics was brief: "Well, what if they're right?"
The practical implication is that Amodei wants government and industry to move beyond minimizing the possibility of AI-driven job loss. He summed up the starting point in a short phrase: "The first step is warn."
What a token tax would try to do
Amodei has also floated a policy response: a "token tax." Under the idea he described, every time a language model generates revenue, three percent would go to the government for redistribution.
He acknowledged that this would not benefit his own company financially. "Obviously, that's not in my economic interest," he said. "But I think that would be a reasonable solution to the problem." If AI grows as expected, he said such a tax could bring in trillions in new government revenue.
Tokens are the smallest units of language that an AI model processes when it generates responses. They can be words, word fragments, or punctuation. In many AI services, the number of tokens used helps determine what customers pay.
That makes a token tax different from a broad corporate tax or a general technology fee. It would be tied directly to revenue from language model usage. In Amodei’s framing, the goal would be to capture some of the value created by AI systems and route it toward people and systems affected by job losses.
He also expects that wider redistribution systems and large-scale, publicly funded retraining programs may be needed if his forecast proves correct. If those forecasts materialize, the labor market would not simply adjust at the edges; it would require a fundamental overhaul.
Preparation would involve companies and lawmakers
Amodei is not only calling for a tax. He also wants business leaders to help employees understand how their roles may change as AI spreads through the workplace.
That matters because the first impact of AI may not always appear as an immediate job cut. Workflows can change, tasks can be reorganized, and expectations can rise as tools become more capable. Employees may need clearer signals from employers about which parts of their jobs are vulnerable and which skills remain important.
He is also calling for better education for lawmakers, saying that most still do not understand how dramatically AI could reshape the economy. His proposals include regular briefings and a congressional committee focused on the social and economic effects of AI at both national and local levels.
The political challenge is that AI policy would need to address uncertainty. If action comes too late, workers may face disruption without support. If policymakers dismiss the warning as hype, they may miss the window to prepare retraining, redistribution, and oversight mechanisms.
The evidence still leaves room for uncertainty
Amodei’s warning is stark, but the source article also notes that the timing and scale of AI’s labor market effects remain unclear. Recent evidence suggests the transformation may not happen overnight.
One study finds that wages and working hours have changed little so far. The Productivity J-Curve theory suggests that the full economic impact of AI may take time to show up because companies must first reorganize their processes around the technology.
Other signals point to both disruption and new work. The World Economic Forum's Future of Jobs Report 2025 says 41 percent of global companies intend to cut jobs because of AI automation, while roles created by emerging technologies may outnumber those lost. The AI Index 2025 says 60 percent of employees believe AI will change their work, and over a third are worried about losing their jobs.
Taken together, the picture is unsettled. Amodei is arguing for preparation before the worst-case labor scenario arrives. The counterweight is that early evidence has not yet shown a sudden collapse in wages or working hours. For workers, employers, and governments, the core question is whether AI job losses arrive gradually enough to manage, or quickly enough to overwhelm existing systems.