AI and jobs: Anthropic study identifies which jobs face AI risk — and which remain largely safe
Anthropic’s research aims to track AI’s effects at an early stage, highlighting potential risks so that policymakers and businesses can act proactively. To support this, they developed a metric called observed exposure, which measures how susceptible different jobs are to AI disruption.
Anthropic’s research introduces “observed exposure,” a new measure of AI displacement risk combining LLM capability and real-world usage.
AI adoption is accelerating, and there is growing concern about job displacement. AI is transforming work across industries, but its effects are uneven. While certain roles may face significant disruption, many others remain relatively safe from automation. Knowing which jobs are at risk can help workers, employers, and policymakers plan ahead.
Anthropic’s research seeks to monitor AI’s impact early, identifying potential risks before major changes occur, allowing policymakers and businesses to respond proactively.
To support this, they developed a metric called observed exposure, which measures how susceptible different jobs are to AI disruption.
Jobs least exposed to AI
Certain roles are difficult for AI to replace because they involve:
-Hands-on human work
-Critical thinking and judgment
-Emergency decision-making
-Care for vulnerable people
-Emotional intelligence
-Adapting to unpredictable conditions
View full Image
Share of job tasks that LLMs could theoretically perform (blue area) and our own job coverage measure derived from usage data (red area). (Credit: https://www.anthropic.com/research/labor-market-impacts)
The occupations with minimal blue or red coverage in Anthropic’s research are the safest from AI.
