Anthropic's Economic Policy Framework

What it argues for

This is Anthropic's proposal for how the United States should prepare workers and the economy for AI-driven job loss, published as the economic companion to its Advanced AI Framework. It opens from the premise that "AI may become a general substitute for human labor" and that "The central challenge isn't how to stimulate growth; it is making sure the gains are widely shared." Its stated values are defensive — "We are not seeking job displacement. We are working to prevent or minimize it" — and its aim is "to buy time—month by month, and year by year—by making the economy more flexible and resilient as AI accelerates." The document rests on three foundations: better measurement, including "reporting requirements for AI labs and firms—including Anthropic—on deployment patterns and workforce effects"; "a small, dedicated unit" of government to track AI's spread sector by sector, modelled on the post-war Council of Economic Advisers; and modernised unemployment-insurance delivery systems. It then sorts responses into three tiers keyed to "the unemployment rate as the primary trigger": at roughly 5% unemployment, universal pre-distributive capital accounts (fundable "with equity in AI companies"), wage insurance, licensing reform, retention tax credits, sectoral training and job matching; at roughly 10%, automatic and uniform UI extensions, sector-specific transition support and basic-needs relief; and beyond historical peaks, income support converging on a common floor plus new tax bases and redistribution mechanisms, on which it says plainly "We are not yet ready to advocate specific policies for this scenario." Two commitments stand out for a company position: at Tier 2 it "would support government regulations and incentives at the firm level that manage the pace of displacement," insisting that "any firm that slows down alone simply cedes the market to those that do not"; and on taxation it says "we are ready and willing to pay our fair share," because "if AI companies generate transformative returns, they have an obligation to ensure those returns are broadly shared."

Stated positions (13)

• Frames the problem as distribution, not growth: "The central challenge isn't how to stimulate growth; it is making sure the gains are widely shared."
• Governments need "expanded investment in our statistical agencies, sustained funding for AI-specific survey instruments, and reporting requirements for AI labs and firms—including Anthropic—on deployment patterns and workforce effects," focused on observable metrics such as AI usage rates, intensity of use and productivity by level of AI adoption.
• Governments should create "a small, dedicated unit" to track how AI is moving through the economy sector by sector, stress-test the institutions that must absorb the shock and "flag the early signals that should trigger a policy response" — an AI equivalent of the Council of Economic Advisers.
• Unemployment Insurance delivery "is not sufficiently prepared for a large labor market shock"; states' legacy UI systems should be modernised, and countries should "invest now in the rails" to expand social-protection coverage quickly.
• Responses should be tiered by severity, "using the unemployment rate as the primary trigger," while also watching labour force participation, underemployment, wages and labour's share of national income.
• Tier 1 (~5% unemployment): "Establish universal pre-distributive capital accounts" — make the federal accounts seeded at birth permanent, extend them to young adults and AI-exposed workers, allow withdrawals for retraining and relocation, and fund them in part "with equity in AI companies"; plus wage insurance, occupational licensing reform, retention tax credits for firms that redeploy workers, employer-connected training grants and job-matching infrastructure.
• Tier 2 (~10% unemployment): "We should consider reforming the trigger, so that extension is automatic and uniform across states" for UI Extended Benefits, alongside sector-specific transition support and monthly basic-needs relief for those who have exhausted UI.
• Supports firm-level pacing of automation if disruption outruns adaptation: "we would support government regulations and incentives at the firm level that manage the pace of displacement. Uniform application across the field is essential: any firm that slows down alone simply cedes the market to those that do not."
• Tier 3 (unemployment beyond historical peaks): extend UI automatically as conditions deteriorate, with benefit levels that "converge toward a common floor rather than continuing to track prior wages indefinitely."
• Names but does not endorse new revenue sources — higher capital gains tax, broad consumption taxes, "sector-specific levies on AI use (measured by tokens, compute, or revenue)," and digital dividends — and redistribution tools including universal basic income and AI sovereign wealth funds: "We are not yet ready to advocate specific policies for this scenario."
• Accepts a tax burden in principle: "Regardless of the tax base or distribution mechanism: we are ready and willing to pay our fair share."
• Places duties on firms as well as government: companies adopting AI should build training into deployment and "decide in advance how freed-up capacity will be used, rather than defaulting to headcount reductions."
• US-focused but meant to travel: "the principles underpinning this framework are intended to be global," though the right policy mix will turn on each country's labour law and social contract.

About this document

A 14-page PDF titled "Anthropic's Economic Policy Framework" and dated June 2026, hosted on Anthropic's CDN and linked from the "Policy on the AI Exponential" page alongside the Advanced AI Framework. The running header gives it a second name, "A policy framework for AI's impact on work". No author is named and nobody signs it; it is written in the corporate first person. Thirty numbered footnotes cite economics and government sources — Brynjolfsson et al., Acemoglu, Korinek and Stiglitz, GAO, BLS, the ILO — plus two Anthropic Economic Index reports. After a one-page opening on the labour-market evidence, it runs through "Our values", "About this framework" and "The foundation" (measurement, a government analytical unit, delivery infrastructure). "The policy framework" then sets out three tiers, each with a lead measure and an intervention table. Tier 1, "Baseline economy (~5% unemployment with churn)", leads with universal pre-distributive capital accounts. Tier 2, "Recession-level disruption (~10% unemployment)", leads with expanding unemployment insurance. Tier 3, "Transformative disruption (unemployment exceeding historical peaks)", covers income through existing systems, candidate new tax bases and redistribution mechanisms, and public investment in human- and community-facing work. The asks are addressed to US federal and state policymakers, with a short passage on global social-protection coverage and a paragraph of steps it says firms, including Anthropic, can take. It closes with a narrow commitment: to be specific at each step, to fund the research that tests whether the measures work, and to revise as evidence changes.

How this sits against AI law

Each stance compared with what EU and US instruments actually require. Where no instrument addresses a theme, that gap is shown rather than hidden.

Almost none of this has a counterpart in AI law on either side of the Atlantic, because AI law regulates the technology and this document is about the labour market that absorbs it. The EU AI Act treats AI used to hire, promote, dismiss and allocate work as high-risk and will make employers inform workers before such systems go in, but it says nothing about how fast AI may displace jobs or who pays when it does; unemployment insurance, benefits and tax stay with Member States. In the US, the nearest match is the workforce chapter of America's AI Action Plan, which covers Anthropic's Tier 1 asks on measurement, a Labor Department research hub and retraining. It is silent on Tiers 2 and 3, and its deregulatory premise runs against Anthropic's willingness to see firm-level rules on the pace of displacement. The only federal vehicle for the kind of AI levy Anthropic lists is a House bill that has not moved beyond introduction.

Source

Date on the page:

June 2026

Source checked:

opened and confirmed on 2026-09-29