In June 2026, Anthropic co-founder and CEO Dario Amodei issued a serious warning in a policy article: The large-scale job loss caused by artificial intelligence is not a short-term economic adjustment or business operation problem, but a structural inevitable result determined by the mechanism of AI technology itself that reproduces human cognitive capabilities. He called on the government and society to face up to this long-term challenge and build an institutional framework to "mitigate shocks and share dividends" in advance.
Unemployment is not a side effect, but AI's "design characteristics"
Amodi pointed out that many technology executives and policymakers currently tend to The employment impact caused by AI is described as "transitional pain"-the natural process of old jobs disappearing and new jobs emerging. But he believes the narrative is overly optimistic and even misleading.
"If the original intention of the AI system is to undertake more mental work, then the unemployment problem it brings is not an accident, but an inherent attribute of the technology's role."
He reiterated his previous prediction:In the next five years, about half of junior white-collar jobs may be replaced by AI, and the overall industry unemployment rate may rise to 10%-20%. These positions include highly structured and repeatable cognitive work such as basic copywriting, primary programming, customer service, administrative support, and financial entry.
Amodi emphasized that this is not a businessThe "abuse of AI" is caused by the inevitable by-product of AI's completion of its core mission-extending and scaling human intelligence.
Response Framework: Mitigation of shocks + share dividends
Faced with this structural challenge, Amodi proposed a dual-track response strategy:
1. Short-term: Mitigating the impact
Strengthen monitoring and data statistics: Establish an official tracking system for the impact of AI on the labor market;
Salary guarantee mechanism: Provide income subsidies for workers who are forced to transfer jobs and reduce their wages due to AI shocks;
Incentives for enterprises to stay in jobs: Encourage enterprises to retain manpower through tax incentives rather than simply replacing them;
Matching skills retraining and employment: Building an efficient career transformation support system.
2. Long-term: Sharing dividends
if AI leads to a decline in permanent demand for human labor, which requires more radical social protection reforms:
Taxing AI companies and setting up special funds;
Increase capital gains tax and adjust the distribution of technology dividends;
Promote universal basic income (UBI) to ensure basic living;
Establish a national capital account so that the whole society can jointly hold the wealth created by AI.
"Our goal is not to help companies reduce labor costs, but to help them develop new revenue and revitalize existing employee production capacity." --Dario Amodi
A subtle shift in the industry narrative
It is worth noting that Amodi's statement comes at a time when the world AI giants collectively turn to the node of "optimistic narrative". Many industry leaders, including OpenAI CEO Sam Altman, have recently emphasized that AI will create new careers, increase productivity, and bring economic growth.
Analysts point out that this shift coincides with many AI companies prepare for IPOs, and capital markets prefer "growth stories" rather than "unemployment warnings." But Ahmodi chose to reiterate the risks at this time, demonstrating Anthropic's careful balance between commercial expansion and social responsibility.
He made it clear:"Even if the prospects of AI are extremely broad, we must plan in advance to protect the rights and interests of those groups who cannot naturally share the dividends of technology."
Amodi's warning tore open A real rift behind the AI craze: When technology can replicate human beings 'core asset-cognitive capabilities-at extremely low cost, the traditional economic logic of "labor for income" will face fundamental challenges. Instead of fantasizing about "automatic job regeneration", it is better to face up to the institutional innovation of "how to redistribute value." The ultimate test of AI may not be in chips or algorithms, but in whether we can build a social contract where the more advanced technology is, the more benefits most people will benefit.