The first AI casualty may be entry-level hiring: Here's what new Anthropic study found

Artificial intelligence has not yet led to widespread job losses, but early evidence suggests hiring may be slowing for younger workers in occupations most exposed to the technology, according to new research by AI firm Anthropic.

In a report released on Thursday, the company said unemployment rates in highly AI-exposed professions have remained largely unchanged since late 2022, when generative AI tools such as ChatGPT began spreading rapidly.

However, the research detected evidence that entry into those occupations may be becoming harder for young workers.

Using US labour survey data, researchers found that job-finding rates for workers aged 22 to 25 entering highly AI-exposed occupations have declined since 2024.

“AI is far from reaching its theoretical capability,” the report said, noting that the actual use of large language models (LLMs) in workplaces still covers only a fraction of tasks they could potentially perform.

The study introduces a new metric called “observed exposure”, which combines theoretical AI capability with real-world usage data to estimate how much different jobs are being affected by AI tools.

The company identified computer programmers, customer service representatives and data entry workers among the occupations most exposed to AI automation.

Despite that exposure, the report found no systematic rise in unemployment among workers in the most exposed jobs since generative AI tools became widely available.

The analysis compared unemployment trends among workers in the most AI-exposed occupations with those in jobs with little or no exposure. The gap between the two groups has remained largely unchanged since 2022, suggesting that AI has not yet displaced large numbers of workers.

The more visible change has been in hiring patterns.

Data from US labour surveys suggest the rate at which young workers enter highly exposed professions has fallen by about 14% since 2022, indicating that companies may be reducing new hiring rather than cutting existing jobs.

The researchers said the slowdown could reflect employers adopting AI tools that automate routine tasks traditionally performed by entry-level workers.

The study also found that workers in the most AI-exposed occupations tend to be more educated and higher paid, and are more likely to be older and female compared with those in jobs currently unaffected by AI.

The analysis showed that employment in highly exposed occupations is projected to grow more slowly over the next decade.

For every 10 percentage-point increase in AI exposure, projected job growth declines by about 0.6 percentage points, according to comparisons with forecasts from the US labour department.

The report cautioned that the economic effects of AI may unfold gradually, more like the spread of the internet than the sudden disruption seen during the COVID-19 pandemic.

“The effects may not be immediately clear from aggregate unemployment data; factors like trade policy and the business cycle could cloud interpretations of trend lines,” the researchers said.