By Elizabeth O'Brien

Artificial intelligence can perform many tasks. But if it winds up costing millions of Americans their jobs, it wouldn't just leave a trail of unemployment; it could also weaken Social Security.

So far, AI hasn't led to widespread job losses or big productivity gains. But the technology has vast potential to both uplift and disrupt the workforce, with ripple effects for Social Security.

One possible scenario is that more jobs are lost than created, hurting the program's long-term financial health. An estimated 2.5% of U.S. employment is at risk if AI expands across the economy and cuts employment in proportion to productivity gains, according to Goldman Sachs.

A shrinking workforce would be the last thing Social Security needs.

For several decades, Social Security collected more in payroll taxes than it paid out in benefits, amassing a surplus that was invested in Treasury securities. Now, Social Security pays out more in benefits than it collects from the 12.4% payroll tax that's split between employers and employees. (Only wages up to the cap of $184,500 for 2026 are subject to the tax.)

Social Security has been dipping into its trust fund's reserves to make recipients whole, and this steady erosion will cause the retirement fund to be depleted in 2033, according to the Trustees' projections. Barring a Congressional fix, beneficiaries will then face an automatic benefit cut of 23%--the portion that the trust fund had been making up.

How might AI alter this timeline? If the technology causes widespread job losses, reducing payroll tax revenue, the reserves could run out sooner. Conversely, if AI boosts productivity and wages while the workforce continues to grow, that could push the depletion date out further.

The Social Security Administration did not address AI in its most recent Trustees' report; an agency spokesperson told Barron's that the Trustees don't make an explicit assumption about the effect of future technological advancements on the economy, but one is implicitly reflected in productivity assumptions in their report.

The tech industry argues AI's economic gains will outweigh the negatives. Nvidia CEO Jensen Huang -- known as the "godfather of AI" -- said this summer on Fox News that rather than simply enabling more leisure time, as some skeptics assert, AI will enable workers to do more and "we will be busier."

Yet the technology's impact may not show up for years. And some economists see the productivity gains having only a marginally positive impact on Social Security.

Researchers at the Penn Wharton Budget Model at the University of Pennsylvania, for instance, forecast a 4.2% shortfall in the program's revenues over the next 75 years, based on expected payroll taxes. Yet if the economy sees larger-than-expected AI-related wage gains, that would close the deficit by just 4% over the Trustees' 75-year planning horizon.

AI will boost annual productivity growth the most in the early 2030s, but the permanent gain will be less than 0.04 percentage points due to sectoral shifts, Penn Wharton researchers wrote in another analysis.

"The AI effect is not nearly as big as some people would believe," says Kent Smetters, faculty director of the Penn Wharton Budget Model.

Others worry it could be far worse. Even if AI development stopped today, innovations haven't been fully incorporated into the workforce, and that process will take some time, says Lily Vittayarukskul, CEO and co-founder of Waterlily, a platform that uses AI to help consumers plan for future long-term care needs.

"Give it three to five years, and there will be what we expect to be significant job displacement," says Vittayarukskul, who projects white collar job loss of between 5% and 15% in that timeframe.

While the magnitude of AI's impact on the labor force remains a question, some workers and industries will be hit more than others.

Occupations highly vulnerable to AI include administrative and back-office jobs, sales, management, and legal positions, according to Penn Wharton. Least vulnerable are manual work occupations like building maintenance, construction, farming, and repair.

Andrew Biggs, senior fellow at the American Enterprise Institute, doesn't worry that it will put him out of business, saying that AI has reduced the time he spends on once-laborious tasks like coding.

"For me, it's productivity enhancing," says Biggs, who used to work at the Social Security Administration. AI is more likely to displace workers in support roles, he says, such as research assistant.

Ideally, for Social Security, AI would extend Americas' work lives with a positive net effect on payroll tax revenue. The next stage, though, may be outsourcing work to AI-enabled robots (which isn't far-fetched in a decade or two).

"The more AI is removing large tasks, obviously that affects the workforce size," says Jen Burdick, an attorney at Community Legal Services of Philadelphia, helping people claim Social Security disability benefits. "It's not just that we're going to have robot waiters, it also means that people aren't paying into the trust fund."

Write to Elizabeth O'Brien at elizabeth.obrien@barrons.com

This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

December 26, 2025 01:30 ET (06:30 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.