Boston College Study Finds Older Workers in AI-Exposed Jobs Are Leaving Work Faster Since ChatGPT Launched

The Study

Geoffrey Sanzenbacher, an economics professor who authored new research for the Center for Retirement Research at Boston College, found that workers aged 55 and older in AI-exposed industries are leaving their jobs at higher rates than before, according to CNBC. The departures split roughly evenly between layoffs and voluntary exits.

"It's a statistically significant effect," Sanzenbacher told CNBC. "For some occupations, it can be quite large."

Sanzenbacher built the research on Current Population Survey data combined with AI exposure scores from Tufts University's Digital Planet initiative, which measures how much of an occupation's tasks AI can perform. He compared job transitions before and after OpenAI's ChatGPT launched in late 2022.

The finding: before ChatGPT, older workers in AI-exposed roles were significantly less likely to leave their jobs than workers in less-exposed fields. After the launch, that flipped. Those same workers became somewhat more likely to exit, including into unemployment.

Who's Actually Getting Hit

The older workers most exposed to AI disruption tend to be white, far more likely to hold a college degree, and higher-earning than workers in low-exposure jobs, according to the research.

That matters because the standard policy conversation around Social Security and retirement age assumes it's construction workers, warehouse staff, and other physically demanding blue-collar jobs that force early retirement. Sanzenbacher's data suggests AI may be doing the opposite: narrowing the gap in career length between low-paying and high-paying jobs by pushing white-collar professionals out sooner than expected.

"AI exposure may reduce the gap in career length between low- and high-paying jobs," Sanzenbacher wrote, according to CNBC.

Three Possible Explanations, Not One Proven Cause

Sanzenbacher lays out three ways AI could be reshaping how long older workers stay employed. This is a statistical correlation drawn from survey data, not a controlled experiment proving AI directly caused any individual's job loss.

First, automation could simply be replacing older workers outright, pushing them into unemployment or out of the labor force entirely. Second, the pressure to learn and adopt new AI tools might be driving some older workers to seek different jobs without AI exposure, or to retire rather than retrain. Third, and this cuts against the doom narrative, generative AI could make some workers more productive, raising wages and letting people stay in the workforce longer doing more engaging tasks.

CNBC's coverage presents all three possibilities fairly, but doesn't resolve which mechanism is dominant. The research shows older workers in AI-exposed fields are leaving jobs more. Whether that's mostly forced obsolescence or mostly voluntary reinvention isn't settled by this paper.

Why This Matters for Social Security

The timing lines up with a bigger fiscal problem. Social Security's trustees reported the program's trust fund could run out in late 2032, according to the trustees' latest annual report cited by CNBC. Washington has floated raising the retirement age for years as one fix.

But if AI is pushing higher-earning, college-educated workers out of the labor force earlier than expected, a blanket increase in the retirement age could backfire on exactly the workers policymakers assumed had the easiest path to working longer. Sanzenbacher's own conclusion is that policymakers weighing retirement-age changes need to account for AI's uneven effects across income and education levels, not treat all older workers as interchangeable.

Nobody in this research is claiming AI is destroying the labor market wholesale. The effect Sanzenbacher measured is real but occupation-specific, and the paper doesn't quantify how many total older workers have left jobs because of AI versus for unrelated reasons like health, caregiving, or ordinary retirement timing.

The open question now is whether this is a temporary adjustment as companies test new tools, or the start of a structural shift that shows up in Social Security's own numbers over the next several years.