Anthropic has introduced an AI Exposure Index, a measure of how much of the work in each white-collar occupation could in principle be performed by large language models. The company released the index on March 5, 2026, alongside an analysis by its economists of whether that exposure is yet showing up in employment data.
The index draws on Anthropic’s own data about how its Claude models are used, mapped against the tasks that make up different occupations as defined in U.S. labor statistics. Occupations where a large share of tasks match things Claude is already doing for users score as highly exposed.
Programmers at the top
Computer programmers rank among the most exposed occupations in the analysis. According to Anthropic, roughly 75 percent of the tasks that make up the job are ones that AI systems can perform today, reflecting how heavily developers already use models to write, debug and review code. Other high-ranking occupations include roles in finance, data analysis, writing and customer support, where the work consists largely of producing and processing text.
Anthropic’s economists stress that exposure is not the same as displacement. A task that a model can perform may still be done by a human for reasons of accountability, quality control or integration with other work, and many jobs include tasks that fall well outside what a chatbot can do.
No clear sign of mass job losses
The company’s analysis found no clear evidence that AI has produced large-scale unemployment in the highly exposed occupations so far. Employment levels in those fields have not diverged sharply from less-exposed fields in the way a wave of automation-driven layoffs would suggest.
The report did identify one early warning sign. Hiring of workers aged 22 to 25 appears to be slowing in the occupations with the highest exposure scores, the economists said. Entry-level roles often consist of exactly the routine tasks that models handle well, so new graduates may be the first group to feel pressure, even if experienced workers are not yet affected.
Why Anthropic is publishing this
Anthropic has positioned the index as a monitoring tool that it plans to update over time, so that policymakers and researchers can watch for displacement effects as they emerge rather than after the fact. The company’s chief executive, Dario Amodei, has repeatedly warned that AI could eliminate a significant share of entry-level white-collar jobs, which makes the current finding of limited impact notable coming from Anthropic itself.
The analysis has limits. It is based on usage of one company’s products, so it captures how Claude is used rather than how all AI tools are used, and the exposure scores describe technical feasibility rather than business decisions. Whether AI ultimately eliminates roles or reshapes them into human-and-model workflows remains an open question that the data so far does not settle.
