Anthropic co founder Jack Clark warned that artificial intelligence (AI) could push unemployment to recession levels even as economic output climbs, in an interview published Wednesday.

Clark, who leads policy at the company behind the Claude chatbot, told Reason’s interview podcast that governments should plan for outcomes that break from past patterns. He said AI might deliver gross domestic product growth well above trend while throwing large numbers of people out of work at the same time, a pairing he urged officials to model now.

The warning carries weight because Clark sits close to the technology’s development and advises governments on its effects. He cautioned that forecasters who predict the exact shape of the future economy often look foolish later, and he framed the joblessness risk as one scenario among several rather than a firm call.

Clark pointed to “a spike in unemployment that you typically only see during a recession” as the kind of case planners should prepare for.

He also expects AI to reach recursive self improvement, the point at which systems design and train their successors without human direction. Clark said that capability could arrive this decade and bet on 2028 if pressed. He argued the progress felt over the past five or six years could compress into the next two or three, then compress again.

On hiring, Clark said current data does not yet show AI erasing white collar jobs, partly because the pandemic distorted employment figures. He noted some softness in early graduate hiring. Within Anthropic, he said, the firm is recruiting more experienced staff, because senior judgment now carries greater value once routine experiments run automatically.

Clark broke with a view, associated with venture capitalist Marc Andreessen, that the Biden administration was hostile to AI development and pushed parts of Silicon Valley toward Trump. That had not been his experience, Clark said, adding that he saw broad continuity across recent administrations on supply chains and model testing.

The interview follows a public clash between Anthropic and the Trump administration. In mid June, Commerce Secretary Howard Lutnick notified chief executive Dario Amodei that the company’s two most powerful models, Fable 5 and Mythos 5, would face export controls barring access by any foreign national. Anthropic disabled the models for all customers and called the directive a misunderstanding tied to a narrow security flaw. Dean Ball, an AI policy specialist who briefly served in the Trump administration, publicly questioned the decision.

Clark said he could not discuss specifics of the dispute, which remains live, but described daily talks with officials over how to export the models without releasing their most sensitive security capabilities.