The latest

Anthropic has released a model charting paths for US growth, wages and employment through 2030, with its most extreme scenario combining 15% annual growth and double-digit unemployment among knowledge workers. Co-founder Jack Clark, who leads the Anthropic Institute, said the scenarios are not predictions and carry no assigned probabilities. Instead, the tool allows users to vary assumptions and examine outcomes ranging from AI behaving like a conventional technology to sweeping economic transformation.

Details

• Model assumptions: Four variables drive the results: the share of economic tasks affected by AI, the technology’s adoption rate, productivity gains per task, and the balance between automating labour and augmenting workers. They also depend on decisions across business, markets, regulation, politics and society.
• Capability shifts: Clark linked the extreme case to sudden advances that could accelerate AI diffusion. He cited a shift in coding capabilities during 2025 and another in cybersecurity in 2026, saying systems may progress from handling individual tasks to completing bundles representing most of a job.
• Employment effects: The substantial and modest scenarios produce more conventional results, with negligible or minor employment effects. Clark said labour-market changes take longer to emerge and often accompany broader, unpredictable economic events, such as recessions, when businesses move towards new staffing norms.
• Smaller companies: AI appears to be supporting competitive businesses with few employees, high operating expenditure and heavy use of computing tokens, Clark said. He pointed to observations from Stripe and others, showing firms combining small workforces with extensive AI consumption.
• Physical constraints: Productivity gains would not automatically remove political, regulatory or social bottlenecks. Clark acknowledged that AI might automate infrastructure designs, permits or parts of clinical trials, while whether those capabilities translate into construction or medical progress still depends on policy and implementation.
• New industries: Clark cited former Federal Reserve chair Ben Bernanke, a member of Anthropic’s Long-Term Benefit Trust, who argued that GDP growth above 5% would require entirely new industries. Such an economy would involve new products and services, and broader changes in how the world operates.
• Sceptical view: Financial Times columnist Sarah O’Connor questioned whether 15% growth could coincide with unemployment approaching one-fifth of cognitive workers and steep wage declines for those remaining employed. She argued that severe social and political disruption could emerge before the economy reached that point.

Background

The Anthropic Institute was established to bring assumptions and information held inside frontier AI laboratories into public debate. Its model spans relatively normal outcomes and extreme cases intended to clarify the choices governments, companies and societies could face under different paths of AI development and adoption.

What’s next

The central test comes in 2030, when US GDP growth, knowledge-worker unemployment and wages can be compared with the modelled outcomes. Earlier indicators include AI adoption, task-level productivity and employment changes linked to coding and cybersecurity tools.