US vs Cuba: AI Job Risk Across the Florida Straits

The closest rivals: scores that mislead

At first glance, the US-Cuba AI exposure comparison looks surprisingly close. The US average is 5.07/10 and Cuba's is 4.24/10 - a gap of 0.83 points on a 10-point scale. That is smaller than the US-Iran gap (1.36 points), smaller than the US-Mexico gap (1.24 points). Both countries peak at 8.5/10 for clerical workers. Both score 6.5/10 for professionals. The task-level susceptibility framework produces almost equivalent readings for many occupation groups.

This apparent closeness is analytically misleading. AI exposure scores measure whether the core tasks of a role fall within current AI capability. They do not measure deployment speed, infrastructure access, economic incentive, or political structure. On every one of those dimensions, the US and Cuba diverge as sharply as any two countries on Earth.

The US Bureau of Labor Statistics Occupational Employment and Wage Statistics (OEWS) May 2025 release covers 143,066,500 workers. Cuba's most recent ILO ILOSTAT data (CC BY 4.0) is from 2012, covering 4,844,000 workers. A 13-year data gap is significant anywhere. In Cuba's case, it spans the Obama-era partial normalization (2015-2016), the subsequent re-tightening under both Trump and Biden administrations, the 2021 protests and their aftermath, and severe economic contraction driven by COVID-19 tourism collapse, tightened sanctions, and fuel shortages. The 2012 occupational structure almost certainly no longer reflects current Cuban employment distribution accurately.

Side-by-side: occupation groups in both countries

The table below compares both countries across ISCO-08 occupation categories using ILO ILOSTAT data for Cuba (2012 data year - see data disclosure above) and BLS OEWS May 2025 for the US. Cuba wage data is not available at ISCO-1 level in the ILO ILOSTAT series.

Cuba's professional class (748,200 workers, 15.4%) is notably large by global standards for a country at its GDP level. This reflects Cuba's historic investment in education, particularly in medicine and engineering. Cuba has one of the highest physician-to-population ratios in Latin America (World Health Organization data). However, Cuba's professionals face a fundamental constraint: the digital tools, platforms, and infrastructure required for AI-augmented professional work are largely inaccessible or unreliable in Cuba. Cuba's internet penetration is among the lowest in the Western Hemisphere, with connectivity concentrated in state institutions and mobile data constrained by both infrastructure and cost relative to wages.

Why Cuba's risk velocity is 0.5: the state economy constraint

Cuba's economy is estimated at roughly 80% state-owned (International Monetary Fund, Cuba Economic Outlook). This is the structural fact that makes Cuba's AI exposure scores almost theoretical in the near term. In market economies, AI adoption is driven by competitive pressure, labour cost arbitrage, and investor-driven efficiency mandates. When a US financial services firm finds that AI can process claims at lower cost than human clerks, the competitive incentive to deploy is clear and fast. When a Cuban state enterprise faces the same calculation, the decision passes through government planning bodies, is weighed against employment stability mandates, and requires foreign exchange allocation that Cuba's severely constrained hard currency position rarely permits.

The US embargo - technically the Cuban Assets Control Regulations (CACR) administered by the US Treasury's Office of Foreign Assets Control - prohibits most US technology exports to Cuba. This includes the cloud computing services, enterprise AI platforms, and high-performance computing hardware that underpin commercial AI deployment. US AI providers including the major large language model API providers are subject to these restrictions. Cuba cannot purchase them legally, and the workarounds available to other sanctioned economies (routing via third-country intermediaries) are more difficult given Cuba's hard currency shortage and limited correspondent banking relationships.

The professionals paradox: high score, no deployment path

Cuba's 748,200 professionals (15.4% of tracked employment, score 6.5/10) represent an interesting case. Cuban doctors, engineers, and educators are exposed to AI task susceptibility in the same way as professionals in any other country - their diagnostic, analytical, and research tasks fall within AI capability. Cuba's medical education system has produced professionals with skills that are globally competitive; roughly 30,000 Cuban medical personnel work abroad under government-to-government agreements (primarily in Venezuela, Bolivia, and Africa), generating hard currency for the Cuban state.

But exposure in principle does not translate to displacement in practice without the tools. Cuban hospitals and clinics operate with limited diagnostic equipment, inconsistent pharmaceutical supplies, and heavily restricted internet connectivity. The AI-augmented diagnostic tools available at US teaching hospitals are not accessible in Cuban medical facilities. Cuban software engineers face similar constraints: Cuba does not have a meaningful technology startup ecosystem, and the foreign exchange earnings from software exports require state approval and routing through official channels.

The result is a professional class that is structurally exposed to AI task competition on the scoring framework but faces near-zero actual displacement risk in the near term - not because their skills are safe, but because their working environment cannot access the tools that would displace them.

Economy context: US vs Cuba side by side

Cuba's 1.75% unemployment rate (World Bank 2025) requires important context. Cuba maintains a policy of full employment as a state guarantee - the state sector absorbs workers regardless of productive demand. This creates structural underemployment rather than unemployment: workers are employed by state enterprises but not necessarily in roles that match their skills or generate economic value. The low headline unemployment figure does not indicate a labour market operating at capacity; it reflects the state's employment mandate rather than private sector demand for labour.

Cuba's HDI of 0.762 (UNDP HDR 2025, rank 97, 2023 data year) is meaningfully higher than its GDP per capita would predict. Cuba's health and education systems have historically been strong relative to income level - a legacy of heavy public investment in those sectors since 1959. But the HDR 2025 also reflects the pressures Cuba has faced: economic contraction, emigration of educated workers, and declining living standards have eroded some of the gains that made Cuba's HDI an outlier.

The safest jobs from AI in both countries

Cuba's elementary occupations at 834,900 workers (17.2% of the 2012 total) score 2.0/10 - the lowest possible AI exposure. This is the largest single occupation group in Cuba's tracked data. The US elementary occupations (3,846,500 workers, score 2.0/10) also represent the lowest-risk group. Physical tasks in unstructured, variable environments resist automation equally in both economies - though for entirely different reasons. In the US, the resistance is primarily technical (AI systems cannot yet handle the full range of physical variability cost-effectively). In Cuba, it is compounded by the absence of the automated equipment and AI systems that might otherwise replace this work.

What this means for workers in both countries

For US clerical workers, the 8.5/10 score reflects immediate, ongoing deployment of AI in their roles. Financial services, insurance, healthcare administration, and legal support have all seen documented AI-driven workflow changes in 2025 and 2026. The 16.5 million US clerical workers face the highest near-term disruption of any occupation group in either country.

For Cuban clerical workers, the 8.5/10 score is a theoretical exposure measure that has no near-term practical relevance. Cuba's 385,000 clerical workers in the 2012 data are employed by state ministries, hospitals, schools, and enterprises. Their roles are determined by government planning, not market competition. Even if AI tools capable of automating their work became available, the incentive structure for a state employer to deploy them - at the expense of creating unemployment that the state then has to manage - is fundamentally different from a private employer facing competitive pressure.

For US professionals (42.9 million, score 6.5/10), AI augmentation is accelerating across legal, medical, financial, and engineering roles. For Cuban professionals (748,200 in 2012 data), the structural exposure exists but deployment paths do not. Cuban doctors working on government contracts abroad in countries like Venezuela may encounter AI medical tools in those host countries - creating an interesting asymmetry where Cuban labour exported abroad faces higher AI exposure than the same workers would face at home.