Clark cites an internal model for about $44 trillion of US GDP

On Saturday, a Reuters report carried by Channel NewsAsia said Jack Clark told officials at the IMF and World Bank annual meetings in Bangkok that one of Anthropic’s internal models forecasts, under certain conditions, a rise in US economic output of almost 50 percent by 2030. The page’s published time is 09:14 UTC on 17 October 2026. Clark is Anthropic’s co-founder. The same Reuters copy, and Anthropic’s leadership page, call him head of public benefit.

The wire says he put the estimated potential US GDP at $44 trillion, against a current level of just over $30 trillion, and that the figure “sounds very outlandish, I know.” Reuters presents that number as his account of an internal company model. It is not a figure the story attributes to the IMF or the World Bank, and the copy does not present it as a forecast Anthropic has said will happen.

What he said the number assumes

Clark said, in the Reuters copy, that the forecast assumed “the technology continuing to get better. And if GDP were to grow this much, then economic doctrine suggests that new industries and inventions would need to have been created to power such a major economic expansion.”

He said international cooperation was needed to erect “unusual and serious rules,” and that the window for policymakers was shrinking. The copy quotes him on misuse: “Being able to use AI for science opens us up to the misuses of science as well, biological and cyber weapons and things we cannot yet imagine, and we have many choices to make.” Biological research, in the same copy, is the area “where we want the strictest human oversight and transparency to maintain safety.” He said AI was still under human control and that the window for safeguards was now “a small number of years.” On jobs, Reuters reports him saying AI was causing losses in fields with “high AI exposure,” naming call centers and customer service, and enhancing employment in scientific research.

A Reuters photo caption on the page places him at the session “AI and the Global Economy: From Risk to Opportunity,” at the Queen Sirikit National Convention Center in Bangkok, on 16 October 2026. The article text says he spoke on Saturday, 17 October.

A public scenario near the same figure

Anthropic already publishes a scenario tool whose extreme case sits next to the number Clark used. Scenarios for our Economic Future, labeled explorer version 1.0 from September 2026, lists US GDP in 2030 at 2025 prices: $34.1 trillion in a modest scenario (1.6% above a no-AI path), $36.3 trillion in a substantial scenario (8.3%), and $44.4 trillion in an extreme scenario (32.4%). The same page describes today’s US economy as over $30 trillion of value in the past year. A step from just over $30 trillion to about $44 trillion is the “almost 50 percent” comparison in the Reuters copy. The 32.4% on the page is a different comparison: the extreme case against the model’s no-AI path for 2030.

The Reuters copy calls what Clark cited “one of the company’s internal models.” It does not say he named this page. The page’s credits say he provided direction. The page says the future is not predetermined. It also says the explorer leaves out policy responses, business cycles, demand or financial disruptions, and catastrophic risks.

The companion paper, Economic Scenarios for Transformative AI (Anthropic Institute Working Paper No. 2026-02, September 2026), says the scenarios are illustrative and that the authors’ views do not necessarily represent Anthropic. In the extreme case it describes, annual GDP growth rises to 15 percent, the labor share falls from 60 to 45 percent, and nearly one in five cognitive workers is unemployed. Those are scenario outputs.

How this sits next to Anthropic’s other policy notes

A 16 October study of Anthropic’s own workforce uses the company’s hiring records. It is about jobs inside Anthropic, which is a different document from a national GDP scenario. A reported draft bill, which Politico described and Reuters relayed on 16 October, is a pitch to US lawmakers. The bill text is not public. Clark’s Bangkok remarks, in the Reuters copy, ask for international rules. They are a speech.

For a reader, the comparison is the part that changes the sentence. Against today’s just-over-$30 trillion, Reuters has Clark at about $44 trillion and calling it outlandish. Against a no-AI path for 2030, the public extreme scenario is $44.4 trillion, 32.4% higher, one of three cases.