Anthropic builds early warning system to track AI’s impact on white-collar jobs

Anthropic

• Anthropic launches index to track AI's impact on job automation
• Programmers, data entry, and customer service most at risk
• No major job losses, but entry-level hiring may slow

Did our AI summary help?

Artificial intelligence company Anthropic says it is working on an early warning system designed to detect whether AI could eventually disrupt white-collar jobs, according to a report by Axios.

The company, which develops the Claude AI models, has introduced a new index aimed at tracking how exposed different occupations are to automation by large language models. While the initial findings show limited evidence that AI is currently causing job losses, Anthropic says building the system early could help economists detect future disruptions more accurately.

Anthropic economists Maxim Massenkoff and Peter McCrory explained in a new research paper that the goal is to monitor potential economic changes before they become obvious.

“By laying this groundwork now, before meaningful effects have emerged, we hope future findings will more reliably identify economic disruption than post-hoc analyses,” the researchers said.

The effort comes amid an ongoing debate among economists about how to measure the real-world impact of AI on employment. Anthropic CEO Dario Amodei has been among the most vocal tech leaders warning that advanced AI systems could significantly disrupt the job market.

Anthropic’s index works by examining the specific tasks involved in different occupations and estimating which of those tasks can be handled by large language models. It also analyses anonymised data to see which tasks are already being performed by AI today.

Jobs are considered more “exposed” when a large portion of their tasks can be automated and when those tasks are already being handled by AI tools in practice.

According to the company’s analysis, computer programmers are among the most exposed occupations, with AI capable of covering around 75 percent of their typical tasks. Other roles that rank high on the exposure list include customer service representatives, data entry workers and medical record specialists.

However, many jobs remain relatively insulated from AI automation. Anthropic estimates that about 30 percent of occupations do not meet the threshold to be considered exposed. These tend to be more hands-on roles that rely heavily on physical work and human presence, such as cooks, lifeguards and dishwashers.

Despite concerns about automation, the company says there is little evidence so far that workers in highly exposed jobs are experiencing higher unemployment.

Researchers found that the unemployment gap between AI-exposed jobs and AI-resistant jobs has changed only slightly since the release of ChatGPT, and the difference is statistically insignificant.

Still, the data hints at early shifts in hiring patterns. Anthropic says there are signs that hiring for younger workers — particularly those aged 22 to 25 — may be slowing in AI-exposed occupations, suggesting entry-level roles could be among the first to feel the effects of AI adoption.

Looking ahead, the company believes its tracking system could help economists identify subtle changes in the labour market that may otherwise be difficult to detect.

Anthropic researchers note that AI’s current level of automation is still far below its potential, meaning the technology could have a much larger impact on jobs in the future.