I watched an HR manager close her laptop mid-meeting when a layoff email pinged the room. You could feel the future rearrange itself in that silence. It was a moment that felt like a calendar flipping to a new era.
I read Anthropic’s new economic policy framework the way I read a weather report before a hurricane: you don’t want to panic, but you want to know which houses to shore up. You should know the headline: Anthropic has pledged $350 million (€320 million) to study and pilot responses if AI drives mass job disruption.
On an office floor in 2026, conversations are about redeployment and benefits
Anthropic begins from a simple admission: “Some amount of displacement, though we cannot say how much, may be an intrinsic consequence of the technology, and our responsibility is to prepare for it and respond to it.” That sentence hands you the frame for their work. They map three scenarios — roughly 5% unemployment, 10%, and an extreme “unprecedented” case that could look like the Great Depression’s scale — and propose tiered policy responses.
On university campuses, students are asking whether their degrees will still pay rent
The 5% scenario reads like a policy stress test. Anthropic proposes expanding capital accounts seeded at birth and letting young adults draw on them sooner. They want those accounts to hold more than index funds — potentially stakes tied to AI productivity — and to pair savings with workforce training grants, occupational licensing reform, wage insurance, and incentives for companies that retain or redeploy staff.
What would 25% unemployment mean for the economy?
If you push unemployment to a quarter of the workforce, you stop playing around. Anthropic’s “unprecedented unemployment” bucket is where they admit uncertainty and raise the big tools: income replacement at scale, new sources of tax revenue, and broad redistribution mechanisms. Names on the table include universal basic income, sovereign-wealth-style funds funded by AI gains, equity-sharing where workers hold stakes in AI firms, and vastly expanded pre-distributive capital accounts.
In a factory, a foreman once told me machines took the repetitive, then new work followed
The 10% scenario is a middle ground that blends emergency relief with managed pacing. Anthropic prioritizes expanded unemployment insurance supplemented by sector-specific supports and basic-needs relief. They also suggest policymakers might need levers to slow the rollout of disruptive systems, and incentives for firms to phase changes so workers have time to reskill.
Can governments or companies stop AI from causing mass unemployment?
Short answer: not entirely, and that’s the point of Anthropic’s plan — to buy space for adjustment. They argue a mix of incentives and safety rails is necessary: rewards for firms that retain talent, grants for retraining, revised licensing to lower barriers between careers, and wage insurance to bridge gaps. The intent is to keep people attached to work even as job definitions shift.
At a policy dinner in Geneva, officials asked who will pay for all of this
Anthropic says new taxes are inevitable if income replacement becomes widespread. They list options: higher capital gains, broad consumption taxes, sector levies on AI use measured by tokens, compute, or revenue, and “digital dividends” financed by taxes on the digital sector. Each option carries political and economic trade-offs, and Anthropic admits it’s researching what scales fairly.
They’ll also be testing mechanisms: equity-sharing programs, AI sovereign-wealth vehicles that capture productivity gains, and dramatically expanded capital accounts seeded at birth. The company hopes these ideas land on agendas from the G7 to the AI Summit in Geneva.
At a startup demo, founders asked whether models could draft these policies
You might wonder whether Claude or other models helped write the paper — Gizmodo asked Anthropic and found no immediate answer. It would be oddly circular if AI helped design the fixes for AI-caused job disruption, but it would also be practical: models can model scenarios quickly. Sam Altman hinted at that logic years ago when he said a generally intelligent system should be asked to find investment returns for stakeholders; that’s the same reflex — asking systems to plan the economic future they’re reshaping.
Anthropic’s pledge of $350 million (€320 million) buys them runway to model revenue sources and redistribution schemes. They list a menu of possibilities and leave the political choices to governments. From Donald Trump to Bernie Sanders, politicians on the left and right are publicly worried, and that political pressure will shape which of these options gain traction.
I’m not telling you to feel safe. I’m telling you the playbook is emerging: gradual fixes for small shocks, insurance and pacing for medium ones, and radical redistribution if jobs evaporate at scale. The metaphor I keep returning to is this: the labor market is a pressure cooker — and policy has to release steam without blowing the lid off the economy.
Do you trust private firms and democratic institutions to share the gains if the machines take most of the jobs?