Anthropic CEO Proposes AI Profit Tax to Fund Worker Retraining and Support

Anthropic chief executive Dario Amodei has proposed a novel policy idea that would require companies deploying advanced artificial intelligence systems to pay a special tax on profits derived from those systems. The revenue would then flow into a dedicated fund designed to support workers whose jobs are displaced by the technology. According to a recent Fortune article, Amodei outlined this concept during an interview in which he argued that the rapid adoption of AI across industries demands proactive measures to address the economic fallout for affected employees.

Amodei, who co-founded Anthropic after leaving OpenAI, has positioned the proposal as a practical response to forecasts that suggest large language models and related systems could automate tasks currently performed by tens of millions of people. Rather than relying solely on existing social safety nets or hoping that new jobs will spontaneously emerge at the same pace as displacement occurs, the executive suggests a targeted financial mechanism. Companies that generate substantial income from AI products would contribute a percentage of those earnings—potentially in the range of one to five percent, though exact figures remain under discussion—into a national or international displacement fund. This pool of money could finance retraining programs, wage subsidies, relocation assistance, and direct cash transfers to workers who lose their positions because of automation.

The concept draws partial inspiration from historical precedents in which industries have been asked to internalize the societal costs of their innovations. For instance, certain sectors have faced carbon taxes to account for environmental damage, or tobacco companies have paid into settlement funds to cover public health expenses. In the case of AI, Amodei contends that the technology’s productivity gains accrue disproportionately to the firms developing and deploying the models, while the costs of adjustment fall on individual workers and communities. By creating a direct link between AI-driven profits and support for those same workers, the policy aims to distribute benefits more evenly and reduce political backlash against the technology itself.

Critics of the idea have raised several practical concerns. Determining which companies owe the tax and in what amounts could prove administratively complex. AI systems often serve as components within larger products, making it difficult to isolate the precise portion of revenue attributable to the AI element. A customer relationship management platform enhanced by machine learning, for example, might derive only part of its value from the intelligent features. Tax authorities would need clear guidelines on attribution, perhaps based on internal cost accounting or usage metrics, yet such standards could invite disputes and creative compliance strategies from corporations.

Another challenge lies in defining displacement itself. Not every job loss in the coming decade will stem cleanly from AI. Economic downturns, globalization, demographic shifts, and ordinary business decisions will continue to influence employment levels. Policymakers would therefore require sophisticated analytical tools to separate AI-related unemployment from other causes. The Fortune report notes that Amodei acknowledges these measurement difficulties but believes they can be overcome through improved economic modeling and transparent reporting requirements imposed on large technology firms.

Supporters point out that the proposal could encourage more thoughtful deployment of AI rather than blanket automation for its own sake. If executives know that replacing an entire department with software carries an additional financial obligation, they might weigh options more carefully. Some firms could choose to pursue hybrid models in which AI augments human workers instead of replacing them outright. This incentive structure might slow the pace of displacement in certain sectors, giving society additional time to adapt through education reform and workforce development initiatives.

The fund itself would require careful governance to avoid waste or political favoritism. Amodei has suggested an independent board composed of economists, labor representatives, technologists, and community leaders to oversee allocation decisions. Transparency measures, such as public dashboards showing inflows, outflows, and outcomes, could help maintain public trust. Money might be distributed through existing agencies like the Department of Labor or channeled via state-level programs tailored to local industry needs. Rural communities hit by agricultural automation, for instance, might receive different forms of assistance than urban office workers displaced by document-processing AI.

International coordination presents yet another layer of complexity. Advanced AI development remains concentrated in a handful of countries, primarily the United States and China, yet the economic effects will cross borders. A tax applied only in one jurisdiction could encourage companies to relocate operations or route profits through low-tax territories. Amodei has therefore floated the possibility of a multilateral agreement, perhaps negotiated through the G7 or OECD, that establishes minimum standards for AI profit taxation. Such an agreement would mirror efforts already underway to address digital services taxes and minimum corporate tax rates.

Beyond the immediate fiscal mechanics, the proposal touches on deeper questions about the social contract in an era of accelerating technological change. For much of the twentieth century, productivity growth generally translated into higher wages and broader prosperity, even if gains were unevenly shared. Recent decades have seen a decoupling between corporate profits and median worker compensation in many developed economies. AI threatens to widen that gap further unless deliberate policy interventions occur. By suggesting that companies profiting from AI should help shoulder the burden of transition, Amodei is essentially arguing for a new form of corporate responsibility that extends beyond traditional philanthropy or voluntary upskilling programs.

Labor unions have expressed cautious interest in the concept while emphasizing that any such fund must supplement—not replace—strong worker protections, collective bargaining rights, and minimum wage standards. Some union leaders worry that framing the issue primarily around displacement could distract from the need to ensure quality job creation in emerging fields such as AI maintenance, data curation, and ethical oversight. Others see the tax as a useful bargaining chip that could be used to secure commitments from employers to provide advance notice of automation plans and priority access to retraining.

Economists remain divided on the likely effectiveness of the approach. Optimists believe that AI will ultimately create more jobs than it destroys, citing historical examples from the industrial revolution through the computer age. Pessimists counter that the speed and breadth of current AI capabilities differ qualitatively from earlier waves of automation. Generative models can now perform cognitive tasks that once seemed uniquely human, potentially affecting white-collar professions that previously enjoyed relative immunity. In this environment, a dedicated revenue stream for adjustment assistance could serve as a necessary buffer while societies figure out new ways to organize work and distribute resources.

Amodei has stressed that his suggestion represents one tool among many rather than a complete solution. He advocates complementary policies including expanded access to higher education, portable benefits for gig workers, and public investment in scientific research that could open entirely new industries. The tax proposal, in his view, simply ensures that those who capture the largest financial upside from AI contribute proportionally to mitigating its downsides. Without such contributions, public support for the technology could erode, leading to regulatory restrictions that stifle innovation altogether.

Implementation would likely begin with large language model providers and companies deploying AI at scale, then expand as capabilities spread. Early revenue might remain modest while adoption is limited, but projections suggest that if AI adds trillions of dollars in economic value over the next decade, even a small tax rate could generate billions annually for worker support. Those funds could finance everything from community colleges offering AI-related certificates to experimental programs that test universal basic services in regions experiencing rapid technological unemployment.

The conversation sparked by Amodei’s comments reflects a growing recognition across the technology sector that developers cannot focus exclusively on capability improvements while ignoring societal ramifications. Several other industry figures have offered their own policy recommendations, ranging from regulatory sandboxes to government-led AI procurement strategies that prioritize human-centered design. What distinguishes Amodei’s idea is its explicit focus on financial transfers tied directly to profits rather than broad-based taxation or voluntary corporate pledges.

As governments around the world draft legislation to govern artificial intelligence, proposals like this one may influence the direction of debate. Lawmakers seeking to demonstrate concern for working families might find the notion of an AI displacement fund politically appealing, especially if it can be designed without imposing excessive burdens on smaller businesses. At the same time, technology companies will likely lobby for clear definitions, reasonable rates, and sunset provisions that allow the policy to be revisited as the economic effects of AI become clearer.

The coming years will test whether such market-based mechanisms can successfully bridge the gap between technological progress and social stability. Dario Amodei’s suggestion offers one concrete framework for discussion, grounded in the belief that those who benefit most from artificial intelligence should help those who stand to lose the most in the transition. Whether policymakers ultimately adopt, modify, or reject the idea will shape not only the economic landscape of the 2030s but also the public perception of AI as either a broadly shared opportunity or a concentrated source of private wealth. The proposal, regardless of its final form, underscores the urgency of addressing these questions before widespread displacement creates irreversible social strain.