Generative artificial intelligence, since its rapid spread in late 2022, has revived questions about its impact on the labor market. While we feared a wave of job destruction, initial data suggests a more complex dynamic. Far from a generalized increase in unemployment, AI seems rather to be changing the entry points into certain professions, specifically affecting young workers.
A study by Anthropic, published in March 2026, highlights this phenomenon: the global labor market is not experiencing a systematic increase in unemployment in positions exposed to AI. However, for 22-25 year-olds, the picture is different, with a notable slowdown in new hires in these same professions. This divergence raises questions about the professional integration of new generations and the adaptation of skills to a changing technological environment.
Overall unemployment stable despite AI acceleration
Since the end of 2022 and the widespread diffusion of generative artificial intelligence tools, many observers have anticipated a major disruption of the labor market. Yet the Anthropic study, titled "Labor market impacts of AI: A new measure and early evidence," finds no systematic increase in unemployment rates for workers in positions highly exposed to AI [1]. This finding suggests that AI integration has not yet caused massive job destruction resulting in increased unemployment on a global scale. Companies seem rather to be adapting their working methods and tasks within existing positions, rather than proceeding with widespread layoffs.
This result fits within a historical perspective where technological advances, while transforming production methods, have rarely led to persistent mass unemployment in the long term. Nevertheless, the absence of a visible increase in overall unemployment does not mean the absence of impacts, but rather that these impacts manifest in more subtle and differentiated ways, particularly on entry flows into the labor market [2].
Hiring of 22-25 year-olds slows by 14% in exposed professions
Despite the stability of overall unemployment, the Anthropic study reveals a distinct trend for young workers. Entry rates into employment for 22-25 year-olds in professions most exposed to AI have decreased by approximately 14% since 2022 [1]. This age group, often seeking their first significant professional experience, seems to encounter more difficulties integrating into these sectors. Another study, cited by Anthropic and conducted by Brynjolfsson et al. (2025), confirms this 6 to 16% decrease in employment among 22-25 year-olds in exposed professions, attributing this decline primarily to a slowdown in hiring rather than an increase in departures [1].
Maxim Massenkoff and Peter McCrory, from Anthropic, highlight this divergence: "We find no systematic increase in unemployment for highly exposed workers since late 2022, though we find suggestive evidence that hiring of young workers has slowed in exposed occupations" [1]. This hiring slowdown suggests that employers might prefer more experienced profiles, capable of adapting to AI tools, or that certain junior roles are now partially automated or require different skills from entry into the labor market [3].
Measuring AI exposure: beyond theory
To assess AI's impact, the Anthropic study developed a new measure: "observed exposure." This approach goes beyond the simple theoretical capacity of a large language model (LLM) to accelerate a task. It quantifies tasks that are not only achievable by AI, but also actually used in an automated manner in professional contexts [1]. For this, Anthropic analyzed approximately 800 professions in the United States, combining data from the O*NET database (describing professional tasks), internal use of its own tools (Anthropic Economic Index), and task-level exposure estimates from Eloundou et al. (2023) [1].
A profession is considered more exposed if a significant portion of its tasks can be accelerated by AI, if these tasks are frequently automated via APIs or specific usage models, and if they represent a significant part of the overall role [1]. Among the professions identified as most exposed are computer programmers (with 75% coverage), as well as customer service representatives and data entry operators (67% coverage) [1]. Furthermore, the study notes a correlation between AI exposure and Bureau of Labor Statistics (BLS) employment growth projections for 2024-2034: for each 10 percentage point increase in observed exposure, the BLS growth projection decreases by 0.6 percentage points, indicating weaker growth for more exposed jobs [1].
Specific worker profiles and nuanced interpretations
The study also reveals specific characteristics of workers in professions most exposed to AI. These workers are more likely to be older, women, more educated, and better paid [1]. For example, 17.4% of people in the most exposed group have a higher education degree, compared to 4.5% in the non-exposed group [1]. This suggests that AI might initially affect positions requiring a certain level of qualification, but where part of the tasks is now automatable, rather than only low-skilled jobs.
It is necessary to nuance these observations. The hiring slowdown among young people could have several explanations [1]. Young people not hired in these professions might choose to stay in their current jobs, turn to other sectors less exposed to AI, or return to school to acquire new skills [4]. Job transitions and professional integration behaviors can be complex to measure precisely in surveys [1]. AI's impact could manifest through a reduction in junior positions, slower promotions, or wage compression, without necessarily causing massive layoffs [2, 5]. The study itself acknowledges that "this framework is most useful when effects are ambiguous – and could help identify the most vulnerable jobs before displacement becomes visible" [1].
These initial elements emphasize that AI integration into the working world is not a uniform wave of destruction, but rather a targeted transformation. The question remains how educational systems and employment policies can support this transition to ensure equitable integration of younger generations.
Receive the Journal's analyses directly in your inbox.
Also read
China's CO2 emissions stagnant for 21 months: structural peak or cyclical accident?
China's CO2 emissions fell by 0.3% in 2025. It's a modest figure. But it extends a trend that has lasted for 21 consecutive months: since March 2024, emissions from the world's largest polluter have been "stable or declining."
AI and employment: two March 2026 studies reveal -13% automatable job postings and +20% augmented roles
Does AI destroy jobs? Two studies published in March 2026 — one by Harvard Business School, the other by Anthropic — provide the first solid empirical data. And the answer is more nuanced than the public debate suggests.
The malaria vaccine in Nigeria: 200,000 children vaccinated and a 50% drop in cases in Kebbi State
Malaria killed 608,000 people in 2022, 95% of them in sub-Saharan Africa and 78% children under five. And for the first time, a vaccine deployed on a large scale shows measurable results in the field.