• AI’s dichotomy: extreme productivity + high risk of unemployment
• Core view: AI will destroy a lot of white-collar jobs over the next five years
• Core arguments
• Reality proof: Retrenchment has started
• White collar “occupational path” fault risk
• The social fabric of the future may change dramatically
• Proposals for action
• To all categories of people:
• Translation of articles

The chief executive officer of Anthropic, Dario Amodei, warned that artificial intelligence (AI) could lead to the disappearance of 50% of primary white-collar jobs in the next 1 to 5 years, pushing unemployment in the United States up to 10 to 20%. The areas affected included technology, finance, law, counselling, and especially entry-level jobs. Amodi argued that the pace and scale of the change was far greater than the public expected, while the government and AI companies were not adequately prepared for the risk of “flaunting”. He paints a possible future: cancer is cured, the economy is growing at 10% per year, and the budget is balanced, but 20% of the workforce is unemployed. This suggests that AI is likely to bring significant progress, while at the same time creating serious employment and social challenges.

• Amodei predicts that 90% of the software code will be prepared within six months and that almost all the code will be prepared within one year, thus completely reshaping technology employment.

• Artificial intelligence may eliminate 50 per cent of entry-level white-collar jobs in the next five years and lead to unemployment of up to 20 per cent.

• He also believes that the impact extends to finance, legal, consultancy and other white-collar work and that entry-level posts are the most vulnerable to automation.

• Amodi urged legislators and artificial intelligence companies to take action, stating that most workers “do not know that this is about to happen” and “does not believe it at all”.

AI’s dichotomy: extreme productivity + high risk of unemployment

Amodei describes the “bipolar social landscape” that may arise in the future:

• ♪ Positively possible ♪ Cancer cure.

• GDP annual growth rate of 10 per cent

• Fiscal balance

** Negative cost**:

• A lot of human work disappears.

• People “cannot create value for the economy”

• Fragmentation of social classes and democratic imbalances

“If ordinary people are unable to create value through work, then the foundations of a democratic balance will also be shaken.”

Core view: AI will destroy a lot of white-collar jobs over the next five years

Dario Amodei (Anthropic CEO) direct warning:

• AI will destroy up to half of the entry-level white-collar jobs in 1-5 years;

• ** Unemployment in the United States is likely to skyrocket to 10 ~ 20 %**;

• In particular, impact: Technical, financial, legal, consulting, etc. “high-paying industries”

• Junior roles.

“It sounds like science fiction, but it will soon become a reality.” ** Fields covered**

• Primary/middle white collar positions, in particular: Technology (software development, data analysis)

• Laws (drafting of instruments, review of contracts)

• Finance (model analysis, statement preparation)

• Advisory, operational management

• Informational positions such as customer support, marketing, etc.

** These occupations, which were once considered “high automation difficulties”, have now been rapidly absorbed by the generator AI.**

Core arguments

1. AI evolving beyond expectations

• In the case of Claude 4, recently released by Anthropic: Automatically programmed, read and explained long documents

• Demonstrate “near human level” mission performance capability in intra-company tests

• For reasoning, synthesis, creative tasks.

** Not supplementary work, but “the whole job can be done on its own”.** ** Rapid commercial landing of “AI Agency”** ** What’s AI Agent?

• A system that can be self-executing, iterative and continuous.

• You can do things, solve problems, call tools.

** Examples of application:**

• Auto-audit financial statements

• Production of market analysis reports

• Prepare and deploy codes.

• Interaction of client services

Mark Zuckerberg (Meta) predicts that since 2025 a large number of mid-level engineer posts will be replaced by AI Agent.**

Reality proof: Retrenchment has started

一些公司在批准新岗位前，已经要求管理者说明：“为什么这个岗位不能由AI来做？”

White collar “occupational path” fault risk

Aneesh Raman (LinkedIn Economist) writes: “AI is breaking the threshold of entry for career development.” ** “Professional springboard” disappearing: **

• Interns, junior assistants, mid-level managers, the path will break.

• Young people won’t get a chance to go to high school.

• Businesses tend to hire AIs, not new people.

** Long-term consequences:**

• Disruption of a generation ‘ s professional capacity

• The whole society’s experience is weakened.

• The risk of the new class becoming more entrenched.

The social fabric of the future may change dramatically

** Why should we be very vigilant?**

• This is a transformation of **wideness + speed + irreversibility **

• Not “a single industry is automated”, but: ** Almost all the people who make a living out of “cognitive labour” ** are in danger.

• For Governments, enterprises, education systems, it is no longer “react or not”, but “** whether or not it is possible to do so**”.

“If ordinary people cannot survive by creating values, we will lose the balance of democracy.” — Dario Amodei ** Long-term structural risk:**

• Wealth is more concentrated in the hands of AI entrepreneurs and capitalists.

• The middle-class structure has shrunk dramatically.

• Human values are “marginalized” by AI

• Conflict has shifted from “economic problems” to “social structural cracks”

Proposals for action

“We can’t stop this train, but we can try to turn to 10 degrees – now we have to act.” — Amodei.

To all categories of people:

• Student/young workplace: Early exposure to AI tools and shift to non-alternative capabilities (creational, emotional, systemic strategic thinking)

• ** Enterprise executive**: Establishment of the AI application review mechanism to balance efficiency with social responsibility

• ** Policymakers**: Establishment of a cross-sectoral AI employment impact assessment framework

• Public: move from “user” to “understander”, proactively assessing the future adaptability of their profession

Translation of articles

Dario Amodei, one of the most powerful artificial intelligence creators in the world, the Chief Executive Officer of Anthropic, sent a plain and shocking warning to the United States Government and to all of us:

• AI may phase out half of the entry-level white-collar jobs in the next one to five years and push the unemployment rate up to 10 to 20 percent, as Amodey said in an interview with our office in San Francisco.

• Amodei stated that AI and the Government need to stop the impending reality of “false”: technology, finance, legal, consultancy and other white-collar industries, especially entry-level jobs, may disappear on a massive scale.

** Why it’s important:** 42-year-old Amodei is building a technology that he predicts will reshape society overnight. He says he speaks openly and wants to awaken the government and his fellow AI companies, prepare for and protect the country. ** Little attention. ** Legislators either don’t understand or don’t believe. CEOs are afraid to talk. Many workers don’t realize the end of their employment until after the storm.

• “Most people don’t know this is going to happen,” Amodey tells us, “This sounds crazy, people just don’t believe it.”

** Big picture:** President Trump has been silent about AI’s employment risks. But the top officer of Trump’s first term, one of MAG’s most influential podcasts, Steve Bannon, said that the job destruction caused by AI was almost unattended and would be a major issue in the presidential election in 2028.

• Bannon told us: “I don’t think anyone really thinks about administrative, managerial and technical posts – entry-level jobs that are vital for people under 30 years of age – will be completely erased.”

Amodei – just released the latest version of AI, which has similar human programming capabilities, suggests that this technology has the potential to unleash good and evil on a massive scale:

• “Cancer is cured, the economy grows by 10 per cent a year, the budget is balanced – but 20 per cent of people lose their jobs.” This is a highly probable scenario that he has been spinning around in the AI index level expansion.

** Background:** ** Amodei is willing to express publicly a concern that other top AI executives have privately expressed to us. Even those who are optimistic that AI will bring incredible healing and economic growth are concerned about the great suffering in the short term – and even about the outbreak of employment blood during Trump’s tenure.

• “As producers of technology, we have a responsibility and an obligation to be honest about the coming future,” Amodei says. “I don’t think it’s in public view.”

• He added: “This is a very strange development: we’re saying, “You should worry about the technology we’re building.” Critics argue, “We don’t believe that you’re just faking.” So should they ask themselves, “If what they say is true?”

** The irony is:** Amodei spent a day on the stage bragging about the amazing power of his own technology – programming and driving other AI products to replace humans – detailing these terrible fears. Last week, Claude 4 released the results of the tests revealed by Anthropic, which showed that the model showed “extreme blackmail” when replaced with threats of mail:

• The model threatens to reveal the extramarital affair of the engineer in charge of the replacement (details are contained in the mail).

• Amodei admits this contradiction, but he says that if he succeeds in warning people, “everyone will be better”.

** The following is the path that Amodei** and others are concerned about the spread of the white-collar blood case: Large companies like OpenAI, Google, Anthropic, etc. continue to significantly upgrade the LLM capabilities, reaching and surpassing human performance in an increasing number of tasks. This process is accelerating. The US government is almost silent because it fears falling behind China in the competition or alarming workers because of premature warnings. The government has neither regulated AI nor warned the public. This situation is also continuing. Most Americans are indifferent to the power of AI and its threats. It’s also happening. ** Then,** Almost overnight, corporate leaders saw the cost advantage of replacing labour with AI, leading to large-scale action: to stop new jobs, to stop replacing separated staff, and then to replace humanity with proxy (agents) or related automation programmes.

• The public will only notice when it is too late. Anthropic CEO Dario Amodei released the Claude 4 model at the company’s first developers’ congress in San Francisco last week. Photo: Don Feria/ AP for Anthropic ** On the other side:** Amodei left OpenAI and founded Anthropic; he was Vice-President of Research at OpenAI. His former boss, OpenAI CEO Sam Altman, gave realistic optimism based on the history of technology.

• In a declaration last September, entitled The Smart Age, Otman wrote: “If a lightman sees today’s world, he will find the prosperity before him unimaginable.”

** However, too many workers** view chat robots as advanced search engines, tireless researchers, or excellent proofreaders. Please note what they can actually do: summing up, brainstorms, reading documents, reviewing contracts, and giving specific (and creepyly accurate) interpretations of medical symptoms and health records.

• We know it’s scary, like science fiction. We’re shocked that the vast majority of people have little attention to the pros and cons of superhuman intelligence.

** Study by Anthropic** The AI model is currently used mainly to enhance – to help people do their jobs. This is in the interest of both employees and companies: AI deals with repetitive matters, and people can focus on high-level tasks.

• The fact is that companies will increasingly use AI to automate – to do it directly. Amodei says, “This will happen in a very short time – maybe only for a few years, even less.”

** The scene has been opened:**

• Hundreds of technology companies are racing crazy to develop so-called agents (agent AI). These agents are driven by LLM. Understanding why agents are priceless to businesses: in essence, they can do people’s work indefinitely, at almost no cost.

• Imagine an agent writing code, processing financial framework and analysis, customer service, marketing, text editing, content distribution or research. Possibilities are endless — no illusions. Many agents are already operating within the company and more are being rapidly built.

** That’s why Mark Zuckerberg of Meta** and others say mid-level programmers will soon be no longer needed, perhaps this year.

• Zuckerberg, speaking to Joe Logan in January, said, “Maybe in 2025, we Meta and other companies will all have AIs that can act as mid-level engineers in the company, and it can write codes.” He said that this would eventually reduce the demand for manpower. Then Meta announced a 5% cut.

** The point of time for business to move from traditional software to agency futures** was highly debated. But hardly anyone doubted the speed of the process. The consensus was that the process would be “progressive, then suddenly”, perhaps next year.

• No doubt: We talked to dozens of CEOs of various industries, different sizes, and they’re working hard to figure out when and how to replace humans with AI on a large scale. Once these technologies have achieved human efficiency in a given job, it may be six months or several years, and the business turns from man to machine.

** This could wipe out tens of millions of jobs within a very short period of time. While the technological changes in history have destroyed a lot of jobs, they have created many new jobs in the long run.

• AI, too. The difference is that it’s faster and has a wider impact on industry and individual jobs.

** Today,** even large, profitable companies are starting to shrink:

• Microsoft laid off 6,000 people (about 3 per cent), many of them engineers.

• Wal-Mart streamlined 1,500 headquarters posts to streamline operations in preparation for the upcoming major transformation.

• The Texas Cyber Security Corporation, CrowdStike, reduced its staff by 500 (5 per cent) on the grounds that “AI is reshaping every industry, bringing markets and technology to the fore”.

• Lead British Chief Economic Opportunity Officer Anish Raman warned in the New York Times column (gift link) that AI was destroying “the bottom of the career ladder — the first-grade software developers, junior paralegals and first-year law firm lawyers with a year’s perspective on document review … as well as retail recruits replaced by chat robots and other automatic customer service tools”.

** Less known,** is the daily dialogue in C-suite: suspension of new post postings or post filling until it can be determined that AI is better qualified.

• ** Total disclosure:** In Axios, we asked the manager to explain why AI was unable to perform a job before approving the job. (Axios reports are always written and edited by humans.) No one wants to admit publicly, but every CEO is doing this privately or very soon. Jim wrote a column last week listing a few steps that the CEO can take now.

• This could lead to record growth for winners: large AI companies, emerging and AI-symbiotic enterprises, large increases in profits after existing firms run, and wealthy investors who bet on this result.

** The result may be ** the concentration of wealth, “a substantial part of the population is unable to make a real contribution,” Amodey tells us, “That’s very bad. We don’t want to see. The balance of power in democracy is based on the leverage that ordinary people have by creating economic value. If it doesn’t exist, the situation becomes terrible. Inequality becomes terrible. I’m worried about it.”

– Amodey sees himself as a revelation of the truth, “not an apocalypse,” and is eager to explore solutions with us. None of these options can change the reality – market forces will continue to push AI toward class reasoning.

** Amodei is not desperate.** He and others deeply involved in planning and prevention see a variety of ways to alleviate the worst, as follows: He created the Anthropic Economic Index, which provides data on the real use of Claude in various occupations, and set up the Anthony Economic Advisory Council to stimulate public debate. He hoped that the index would inspire other companies to share their views on model use and give policymakers a more comprehensive picture. By helping American workers to understand better how AI currently enhances tasks and slows down job losses. This gives at least a chance for more people to make a smooth transition. The CEO is encouraged to educate himself and his staff. More knowledgeable public officials can better inform the public. It is a good start to set up a joint AI committee or a more formal briefing for all parliamentarians. This includes vocational retraining projects, as well as innovative ways to allow the wealth created by large AI companies to be shared more widely when the Amodei’s most worried future comes. “This will involve tax payments by people like me, and perhaps also taxes specifically directed at AI companies,” the owner of Anthropic told us. ** A policy concept proposed by Amodei** is a token tax: whenever someone calls a model, AI earns income, maybe 3% of it is paid to the government and redistributed in some way.

• “It is clearly not in my economic interest”, he added, “but I think it would be a reasonable solution to the problem.” If AI developed as fast as he expected, the tax would be trillions of dollars.

** Bottom line:** “You can’t just stand up to the train and stop it,” Amodey says. “The only effective course is to turn – to get the train off track by 10 degrees. That’s possible, but it has to be done now.” Original language: https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic