Anthropic CEO says AI could eliminate half of white-collar entry-level jobs and drive unemployment to 20% within the next five years



Anthropic, an AI company with the goal of 'Responsible AI,' is conscious of the impact of the development of AI technology on humanity, such as

by hiring AI welfare researchers to begin to consider whether AI deserves moral consideration. In an interview, Anthropic CEO Dario Amodei predicted that 'in the next one to five years, the introduction of AI could eliminate half of all white-collar entry-level jobs, causing unemployment to soar to 10-20%.'

AI jobs danger: Sleepwalking into a white-collar bloodbath
https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic



Anthropic CEO: AI Poised to Wipe Out 50% of Entry-Level Jobs in Next 5 Years | PCMag
https://www.pcmag.com/news/anthropic-ceo-ai-poised-to-wipe-out-50-of-entry-level-jobs-in-next-5-years

Anthropic held its first developer conference, ' Code with Claude ,' in May 2025, to talk about its achievements so far and its future prospects. At the conference, Claude Opus 4 and Claude Sonnet 4 were also announced , which have been improved from the previous model in terms of coding and inference performance.

Anthropic releases two models of the 'Claude 4' family, with improved coding and inference capabilities from the previous generation - GIGAZINE



About a week after Code with Claude, Anthropic CEO Amodei shared his views on the future of AI in an interview with news media Axios. 'As developers of this technology, we have an obligation and responsibility to be honest about what's going to happen,' Amodei said. 'Most lawmakers and workers don't believe that AI will cause job losses in the future. It sounds ridiculous, and no one believes it,' he said, warning of the impact of AI on human workers.

According to Amodei, AI technology can 'unleash good and evil on a large scale.' By utilizing the capabilities of AI, diseases that have not been cured until now can be treated, and the economy can grow 10% per year and stabilize finances. On the other hand, 20% of people may lose their jobs. This duality is a very possible scenario ahead of the exponential expansion of AI's power, Amodei pointed out.

The unemployment rate in the United States as of April 2025 is 4.2%, but Amodei predicts that the unemployment rate could soar to 10-20% in just one to five years. In an interview with Fox News, Amodei said, 'AI could also automate jobs in finance, consulting and technology, transforming the job market for college graduates, young professionals and mid-career changers,' specifically talking about the types of jobs that AI is likely to replace.



Major AI companies such as OpenAI, Google, and Anthropic continue to significantly improve the capabilities of large-scale language models (LLMs), developing AI that rivals or exceeds human performance in many tasks. However, the US government is concerned about falling behind China in AI development and has not regulated AI or warned the public, so many Americans have little notice that 'AI's power is growing and it could threaten their jobs,' Amodei said. According to Amodei, business leaders can take steps to reduce costs by replacing human workers with AI almost overnight, and it is likely that the public will only become aware of the threat of AI when it is too late.

In fact, Mark Zuckerberg, CEO of Meta, one of the major AI companies, said in January 2025 that 'probably in 2025, not only Meta but other companies working on AI will have AI that can become like a mid-level engineer who can write code in-house.' He mentioned the possibility of AI replacing office workers. Then, about a week after the remark, Meta announced its 'policy to cut 5% of low-performing employees' to the company, although it was not made clear whether it was related to the development of AI.

Meta plans to cut 5% of underperforming employees - GIGAZINE



Amodei cites a 'token tax' as a policy idea to combat the future in which AI takes over human jobs. A token tax is a tax system in which AI companies would pay 3% of their revenue to the government every time they make a profit using an AI model. By redistributing the collected token tax in some way, the losses caused by AI could be compensated for. Although it would be an economic loss for AI companies, Amodei said, 'I think that would be a reasonable solution to the problem.'

Amodei concluded by saying, 'You can't stand in front of a train and stop it. The only thing that can be done is to take control of the train - to steer it in a direction that's 10 degrees different from the direction it was going. You can do that. But you have to do it now,' reiterating the need to control the development of AI.

in Software, Posted by log1e_dh