Former US Treasury Secretary Robert Rubin has warned that the artificial intelligence (AI) investment boom could deliver enormous productivity gains, but may also generate financial and social risks that markets have not fully priced in.

Rubin, who served as US Treasury Secretary during the internet boom of the late 1990s, said he is particularly concerned about "circularity risk" within the AI ecosystem. The term refers to overlapping commitments among suppliers, customers, and investors, such as mutual commitments between chipmakers and software companies.

"Some large AI companies have made enormous commitments, and at the same time, there are a large number of suppliers, many of whom are borrowing based on these commitments. What happens if these companies cannot fulfill these commitments, or if all the borrowing based on these commitments cannot be repaid? That is what is called circularity risk," Rubin said Tuesday in an interview at the Greenwich Economic Forum.

Rubin said some disruption risks are realistic 鈥?for example, one party failing to honor commitments, which then triggers a chain of defaults. He believes this risk is not "close to zero." The former Goldman Sachs co-chairman said such a scenario is genuinely possible.

This year, companies building data centers and developing AI software have issued debt on such a massive scale that many market participants believe it has pushed up global borrowing costs. This week, benchmark government bond yields climbed further, with the US 10-year Treasury yield touching its highest level since 2002. This in turn has raised government debt servicing costs and intensified concerns about fiscal sustainability in the United States, France, and other countries.

When Rubin joined the Bill Clinton administration, markets were already broadly worried about budget deficits. He said markets are now beginning to show anxiety about large-scale borrowing. "I think what is happening now is that the market is beginning to recognize our fiscal situation, and that recognition is starting to affect the market in a way we have not seen for a very long time," he said.

He also pointed to inflation and an overall decline in market confidence in the government's ability to address fiscal problems. "When you face an unfavorable or unstable fiscal situation, it can further affect market confidence in the government's ability to handle the problem," he said. He added that this dynamic could ultimately affect the stock market as well, because uncertainty tends to suppress investment and productivity growth.

Productivity doubts

Rubin pushed back against the view that AI-driven productivity growth can allow the United States to effectively grow its way out of its fiscal troubles. The former Treasury Secretary said that while an acceleration in productivity could indeed boost GDP growth, those gains may come with significant job losses, especially among white-collar workers.

He predicted that "this will have a very severe impact on knowledge workers." He said lawyers, accountants, and television industry professionals could all face the risk of being replaced, which would raise a difficult question: how should people transition into new jobs?

He noted that the United States currently lacks an effective plan to deal with this kind of employment shock. "Can AI bring higher economic growth? Yes 鈥?it very likely can. But I do not think it will solve the fiscal problems we have to solve," he said.

He also questioned whether massive investment in AI will ultimately generate sufficient returns. "Will these investments ultimately pay off, or not? I do not know, but neither does anyone else," he said. "Some of these investments will turn out very well, some companies will perform excellently, and a large number of others will perform poorly."

AI safety concerns

Rubin said his concerns are not limited to markets and the economy, but extend to the security risks posed by increasingly powerful AI systems. He said he considers AI safety and climate change to be "two enormous existential risks," different in nature from any risks humanity has faced before.

Rubin also noted that the United States faces a dual challenge: trying to put safety guardrails around AI while also having to focus on competition with China.

Rubin also said, "The United States is still the best place to invest." But he believes maintaining that advantage depends on whether the US political system can become effective enough to address fiscal, technological, and other long-term challenges.

"Our political system must become effective enough again 鈥?and right now it cannot do that. It does not need to become outstanding, but it must be effective enough to handle the problems we face," Rubin said.