Dario Amodei Calls for AI Company Tax and Universal Basic Income in Sweeping Policy Essay

Anthropic CEO Dario Amodei published an essay on June 11 laying out his most detailed policy framework yet for governing AI's economic impact. The piece, titled "Policy on the AI Exponential" and posted on his personal site, covers five distinct policy domains and proposes concrete measures including mandatory third-party model testing, government authority to block unsafe AI releases, and, if job displacement becomes severe enough, a universal basic income funded by taxes on AI companies.

Jobs, Displacement, and a Graduated Policy Response

The labor-market section of the essay is its most specific. Amodei argues that AI job displacement may be unavoidable because the technology's core capacity, replicating human cognitive processes, is structurally different from prior automation waves that replaced physical labor. Earlier machines displaced farmers and factory workers while creating new categories of knowledge work. An AI that performs cognitive tasks removes the safety valve that prior transitions relied on. Amodei's earlier analysis of AI's impact on knowledge work laid the groundwork for this position; the June 11 essay turns that analysis into a concrete policy recommendation.

His proposed response is graduated. At moderate displacement levels, he calls for better data collection (which he describes as currently inadequate) and pro-employment incentives that slow the pace of substitution. If displacement becomes more severe and permanent, he proposes mechanisms including universal basic income financed either by taxes on companies that deploy AI, including Anthropic itself, or by higher capital gains taxes on the investors who benefit from AI-driven productivity gains. The framing is conditional, not a firm commitment, but naming both funding mechanisms in print is an unusual degree of specificity for a sitting AI CEO.

Mandatory Testing and Government Power Over Releases

On AI safety governance, Amodei endorses mandatory third-party testing of frontier models before deployment, with explicit government authority to block releases that fail safety evaluations. He frames this not as a hypothetical constraint but as necessary architecture for a period when models are approaching capability thresholds that existing regulatory frameworks were not built to handle. The position is more specific than most lab CEOs have been willing to go in public statements, where endorsing "voluntary commitments" is far more common than endorsing government blocking powers.

The endorsement carries an implicit market consequence. A mandatory pre-deployment testing regime for frontier models would impose compliance costs on all developers, but the burden falls disproportionately on companies building the most capable systems. Anthropic occupies that position. Whether intentionally or not, Amodei's proposal would also raise barriers to entry for new competitors and give incumbent labs more time to engage with regulators before each release cycle.

The fifth pillar of the essay addresses geopolitics: a coordinated strategy among democratic nations to maintain AI leadership relative to authoritarian governments, with explicit attention to export controls on both hardware and model weights. This aligns with positions Amodei has expressed in congressional testimony and in his earlier writing, but the essay assembles the argument more systematically than prior statements.

"AI-driven job displacement may be unavoidable, because it may be intrinsic to how the technology replicates human cognition." Dario Amodei, "Policy on the AI Exponential," June 11, 2026

Money Behind the Words

The essay arrived alongside separate Anthropic announcements committing $350 million across two initiatives. The Economic Futures Research Fund receives $200 million to back research trials and policy evaluations on AI's labor-market effects. Claude Corps, announced the same day, receives $150 million to train 1,000 early-career fellows and place them inside nonprofits at $85,000 annual salaries. Fellows spend 12 months in-person at host organizations receiving training in Claude, with Anthropic also providing each host organization a $10,000 grant and free API credits. The two programs together represent Anthropic's largest public commitment to social programs since its founding.

The response from critics was swift. A company whose products are among the leading candidates to displace knowledge workers is proposing to study the problem and train workers for nonprofit roles that are, by design, outside the sectors most exposed to automation. Fortune's headline captured the tension: "Anthropic is worth $965 billion and just hired 1,000 coaches for nonprofits: 'The fox can't guard the henhouse.'" Amodei's essay, to its credit, does not avoid the core contradiction. His framing acknowledges that Anthropic's own products are central to the displacement dynamic he is proposing to address, and he specifically names AI companies as among the entities that should bear the cost of mitigation.

Whether the $350 million commitment is commensurate with the scale of disruption a $47 billion annualized-revenue company could contribute to is a separate question, as is whether the policy proposals in the essay will find legislative traction in the current political environment. What is notable is the directness. No other CEO at an AI lab of Anthropic's scale has publicly named universal basic income as a likely policy necessity in a document this detailed. The essay positions Amodei as the most explicit advocate for structural redistribution among his peer group, which makes it a meaningful artifact regardless of what happens with the policy proposals themselves. Learn more about Anthropic's approach to responsible AI development and its broader impact on the industry.