Anthropic's own model frames CEO Amodei's job forecasts as an outlier scenario

Anthropic has published an economic model that lays out three scenarios for how AI could reshape the US economy through 2030. In the modest scenario, AI's effect resembles the internet's: GDP growth ticks up slightly and wages hold steady. In the middle scenario, economic growth doubles overall, but wages for knowledge workers stagnate, and workers in roles like programming or call center work would need to retrain into jobs such as electrician or nurse, a hard transition that pushes unemployment higher. In the extreme scenario, economic output doubles every 4.5 years, knowledge worker unemployment reaches 17.9 percent, and labor's share of GDP falls from 60 to 45 percent.

Those extreme-scenario figures land close to warnings Anthropic CEO Amodei made in May 2025, when he said up to half of all entry-level office jobs could vanish by 2030 and that unemployment could reach 10 to 20 percent. The Decoder, which reported on the model, points out that Amodei's own past predictions line up with the scenario the model treats as its most extreme, and by extension, its least likely outcome.

In effect, the company's own economic modeling recasts its CEO's bleakest public warnings about AI-driven job losses as an outlier case rather than the expected result. The modest and middle scenarios describe smaller, more gradual shifts: stable wages in the modest case, and a wage-stagnation and retraining squeeze for knowledge workers in the middle case, rather than the sharp unemployment spike that the extreme scenario, and Amodei's own warning, describe.

Key facts

Anthropic published an economic model outlining three scenarios, modest, middle and extreme, for how AI could reshape the US economy through 2030.
In the extreme scenario, economic output doubles every 4.5 years, knowledge worker unemployment hits 17.9 percent, and labor's share of GDP falls from 60 to 45 percent.
In the middle scenario, economic growth doubles overall, but knowledge worker wages stagnate, and workers such as programmers or call center staff would need to retrain into jobs like electrician or nurse.
Amodei warned in May 2025 that up to half of all entry-level office jobs could vanish by 2030, with unemployment reaching 10 to 20 percent, figures that match the extreme scenario.
The Decoder, which reported the story, says Anthropic's own model treats the scenario matching its CEO's warnings as the most unlikely of the three.

Why it matters

Anthropic CEO Amodei has been one of the most prominent voices warning about AI's effect on jobs. In May 2025 he said up to half of all entry-level office jobs could vanish by 2030, with unemployment reaching 10 to 20 percent. Anthropic's own economic model, covering the same period through 2030, lines up with those figures in its extreme scenario rather than its middle or modest case. The middle scenario still describes programmers and call center workers needing to retrain into jobs like electrician or nurse as knowledge worker wages stagnate. The gap between a company's internal modeling and its own CEO's public warnings is the core of this story: it shows how differently the same set of possible AI outcomes can be framed depending on who is speaking and in what setting.

Who it affects

The scenarios center on knowledge workers, with the middle scenario naming programmers and call center workers specifically as needing to retrain into roles like electrician or nurse. The extreme scenario projects knowledge worker unemployment at 17.9 percent and labor's share of GDP falling from 60 to 45 percent, effects that would land hardest on white-collar and entry-level office employees, the same group Amodei's May 2025 warning named directly. Policymakers and workers trying to judge how seriously to weigh AI job-displacement forecasts are a secondary audience for how this comparison gets read.

How to use it

The Decoder does not link to or name the underlying report, so there is no original document to check the methodology against; what is available is the scenario summary itself. The three bands offer a rough way to place a given AI job-loss claim: modest (an internet-like effect, stable wages), middle (overall growth alongside a wage-stagnation and retraining squeeze for knowledge workers), and extreme (17.9 percent knowledge worker unemployment, labor's share of GDP down to 45 percent from 60). Amodei's May 2025 warning, up to half of entry-level office jobs gone by 2030 and unemployment at 10 to 20 percent, falls inside that extreme band, which is useful for judging how mainstream or fringe a specific AI job-loss prediction is within Anthropic's own framework.

How solid is it

The figures come from Anthropic's own economic model, as reported by The Decoder, so the source is the company itself rather than an independent analyst. The article does not name or link to the underlying report and does not describe the model's methodology or assumptions beyond the three scenarios' headline outcomes, so there is no way to verify how the 17.9 percent unemployment figure or the 4.5-year doubling period were derived. Amodei's May 2025 warning is treated in the piece as established background rather than as new reporting.

Risks and caveats

The framing of the extreme scenario as an 'outlier' is The Decoder's own reading, drawn from comparing Amodei's May 2025 numbers to the scenario Anthropic labeled extreme. The modest scenario carries no numbers at all, only 'slight GDP growth and stable wages,' so it cannot be compared to the middle and extreme scenarios on the same footing. The model's mechanism, methodology and underlying assumptions beyond the three scenarios' headline outcomes are not described in the source, and no publication date is given for the model itself, only the through-2030 horizon of its scenarios and the May 2025 date of Amodei's earlier warning.