Anthropic bods rework AI damage yardstick, find scant labor impact

Anthropic economists Maxim Massenkoff and Peter McCrory report that AI is not eliminating as many jobs as experts have predicted.

Take CEO Dario Amodei as one such expert. In January 2026, Amodei revisited and expanded upon his 2025 prediction that "AI could displace half of all entry-level white collar jobs in the next 1-5 years."

"In the end AI will be able to do everything, and we need to grapple with that," he said in his epic post.

Some day, robots may fold clothes and ferry idle citizens about to spend their government-allocated basic income at depopulated pop-up stores. But not today.

At the moment, it appears AI has had almost no impact on those whose jobs are considered to be "exposed" to automation.

"We find no systematic increase in unemployment for highly exposed workers since late 2022, though we find suggestive evidence that hiring of younger workers has slowed in exposed occupations," write Massenkoff and McCrory in their report [PDF] titled "Labor market impacts of AI: A new measure and early evidence."

There are recent examples of layoffs that have been attributed at least in part to the impact of AI, such as Jack Dorsey's decision to axe around 4,000 employees from Block, or about 40 percent of staff. Given Dorsey's lack of success generating revenue at Twitter when he ran the company in its early years and Block's November 2025 earnings miss, there may be other factors in play.

It does appear that software economics will have to be revisited now that sprawling software projects can be reformulated and relicensed with a bit of prompting.

But overall, Anthropic's econ bods aren't ready to sound the AI alarm bell. As they note in the blog post heralding their work, "the track record of past approaches [to AI labor impact forecasting] gives reason for humility."

Massenkoff and McCrory propose a new measurement that they argue can help clarify things: observed exposure.

This is meant to measure how AI is actually being used as opposed to how it might theoretically be used. They concede, "AI is far from reaching its theoretical capability."

When economists looked at the actual economic impact of AI on workers in Denmark about a year ago, they found no effect on jobs or wages. Anthropic's practitioners of the dismal science see the needle moving a bit, if you look really closely.

But armed with this new yardstick of actual rather than theoretical AI usage, the result is ... pretty much the same.

Anthropic's researchers expect that occupations deemed to have higher observed exposure to AI will grow more slowly through 2034 than other jobs based on US Bureau of Labor Statistics data. And if that comes to pass, the most exposed roles are expected to be those filled by older, female, more educated, and higher paid workers. But we're not there yet.

The average change in the unemployment gap between highly exposed workers and those more insulated from AI since the release of ChatGPT, the researchers observe, "is small and insignificant, suggesting that the unemployment rate of the more exposed group has increased slightly but the effect is indistinguishable from zero."

The exception is younger workers, for whom hiring has slowed among exposed occupations. But even this draws a "meh" from Anthropic's researchers, who note that the 14 percent average estimated decline in the job finding rate between the introduction of ChatGPT in 2022 and now "is just barely statistically significant." ®