Anthropic, the company that makes the Claude AI models, just published something unusual for an AI company: a model that tries to price in its own downside. It built an interactive tool that turns a technical economic report into a simple question. Tell us how capable you think AI will get, and we will show you what the US economy looks like in 2030.
The tool runs three scenarios. In the modest one, AI changes the economy about as much as the internet did, real growth, but gradual and unremarkable in the data. In the substantial one, AI does half of all knowledge work by 2030, and GDP grows at twice its normal pace.
In the extreme scenario, AI outperforms humans at nearly every knowledge task and does it without human help. This is the branch that deserves the closest look, because the numbers are large enough to reshape how a business plans staffing and pricing.
GDP reaches 44.4 trillion dollars in this extreme case, doubling roughly every four and a half years. Unemployment among knowledge workers, coders, analysts, paralegals, support staff, climbs to nearly 18 percent, and their wages fall by more than 10 percent. Economy-wide unemployment rises past 11 percent, a level the US has only seen in its worst recessions.
The detail that got less attention is the money math behind fixing it. Anthropic's own report says that keeping knowledge workers' income at pre-AI levels in the extreme scenario would require transferring about 9 percent of GDP to them, a sum on the scale of Social Security and Medicare combined. No government has ever run a transfer that large in response to a single technology.
This is not the first time Anthropic's leadership has said something like this. Its CEO, Dario Amodei, warned last year that AI could wipe out half of entry-level white-collar jobs and push unemployment to between 10 and 20 percent within five years. What is new here is that the company turned that warning into one branch of a model rather than a standalone claim.
The part worth sitting with is not which scenario wins. It is that all three scenarios point the same direction: a bigger share of new wealth goes to the owners of capital, companies, equipment, and AI systems, and a smaller share goes into paychecks. That shift already shows up in the modest case, just less dramatically.
Anthropic also surveyed more than 10,000 Americans, and the average person's expectations already land close to the substantial scenario for 2030: GDP 10 percent higher than it would be without AI, unemployment near 5 percent. Ordinary people are not being asked to imagine a distant future; they are describing something they already expect within about four years.
For anyone running a business, the useful takeaway is not the exact GDP figure. It is the direction: work that depends on drafting, coding, basic analysis, and customer support is the most exposed in every scenario, while physical and hands-on trades gain relative value. Anthropic built this model and also sells the technology driving the scenario, which does not make the numbers wrong, but means the humility built into the tool, calling itself a work in progress that leaves out policy responses and robotics, deserves as much attention as the headline figures.