mediadev library

Sort by:

Direction:

European AI Competitiveness Beyond Frontier Models: Insights From AI Startups in Five Strategic Markets

Nicole Lemke

Berlin: Interface – Tech Analysis and Policy Ideas for Europe (2026), 37 pp.

"AI development is frequently described as a race for frontier models. But considerable economic value will also be generated by AI development and adoption in existing markets. This data brief examines where EU competitiveness stands in the five applied markets where most of the EU’s AI startup a ... ctivity in our dataset takes place – horizontal enterprise software, healthcare and life sciences, manufacturing and supply chain, finance and insurance, and research and development tools. Drawing on data from Dealroom.co covering 24,468 active AI startups launched between 2012 and 2025 across the EU, Switzerland, the UK, and the U.S., we compare geographic patterns at the level of regions, countries, and cities through original analysis [...] Manufacturing and supply chain is the only market where the EU accounts for a larger share of AI startups than the U.S., though our data likely underestimates the number of U.S. startups in general. This suggests the competitiveness gap across applied AI markets may be even larger than our data indicates [...] AI startup activity in the EU is geographically dispersed: 75 percent of analysed EU AI startups are spread across 95 cities, with no single hub dominating any market. Paris, Berlin, Amsterdam, Munich, and Stockholm are the largest hubs but together account for only a quarter of EU AI startups. EU city rankings across markets largely reflect general AI startup hub sizes rather than genuine market specialisation – but specialised niches do begin to emerge at the local level, e.g. Munich in manufacturing and supply chain and Barcelona in healthcare and life sciences. As our data skews towards larger startup hubs, specialisation patterns in smaller cities and regions may not be fully captured. Our findings have important implications for policymakers. Europe's challenges in AI competitiveness extend well beyond frontier models. The manufacturing and supply chain market, a strategic priority for the European Commission, presents grounds for cautious optimism, where existing industrial strengths align with national and regional specialisation signals based on AI startup activity, especially in Germany and Munich. The geographically distributed startup activity in the EU can support AI adoption across diverse regional economies and existing industrial strengths. At the same time, it raises questions about whether European hubs can achieve the scale and specialisation needed to compete globally. The technological diversity of AI also presents competitive opportunities that go beyond market specialisation alone. Whether the emerging specialisation patterns identified in this analysis develop into globally competitive positions will depend on continued access to growth capital, compute resources, and cross-border market integration." (Executive summary, pages 3-4) more

External Link (31)

Atlas de Inteligencia Artificial para el Desarrollo Humano de América Latina y el Caribe
Deep Insights

Gustavo Beliz et al.

Programa de las Naciones Unidas para el Desarrollo (PNUD) (2025), 466 pp.

"La Inteligencia Artificial (IA) está reconfigurando las economías, las sociedades y la gobernanza en todo el mundo. En América Latina y el Caribe, el Atlas de IA para el Desarrollo Humano ofrece una guía estratégica para navegar esta transformación. El Atlas mapea el impacto de la IA en cuatr ... more

Developing an Artificial Intelligence Maturity Assessment Framework for South Africa. First Annual AI Maturity Assessment 2025

Bonn; Eschborn: Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ); FAIR Forward – Artificial Intelligence for All; University of the Western Cape (UWC) (2025), 107 pp.

"The South African AI MAF serves as a foundational tool for guiding the country’s progress towards AI maturity and is tailored to the country’s unique socioeconomic, cultural, and technological realities. It furthermore aligns with broader continental ambitions (i.e. African Union’s 2024 Conti ... nental AI Strategy), which reflects a strong commitment to leveraging AI for sustainable development. This implies the ethical and inclusive development and implementation of AI in ways that address local challenges. In South Africa, GIZ, through the FAIR Forward project, has collaborated closely with the Council for Scientific and Industrial Research (CSIR), the South African Institute of Chartered Accountants (SAICA) and researchers from the University of the Western Cape (UWC) to develop the AI Maturity Assessment Framework (AI MAF) for South Africa. A design-science methodology was applied to develop the AI MAF for South Africa. The approach involved a structured literature review process, co-design workshops with cross-sectoral stakeholders, expert interviews, validation workshops and external expert review. This first version of the AI MAF consists of eight domains: Domain 1: International and Continental Profile as AI Leader Domain 2: An Enabling Regulatory Environment Domain 3: Citizen Engagement With and Empowerment Through AI Domain 4: Education and Workforce Enablement Domain 5: Innovation and Research Domain 6: Growth of the AI Ecosystem Domain 7: Data and Digital Infrastructure Domain 8: Security and Privacy Each domain includes sub-domains with specific indicators. A total of 68 indicators are included in this 2025 instrument, aimed at reflecting the country’s unique socioeconomic, cultural, and technological realities. An assessment instrument (rubric) to measure the level of maturity of these eight domains was adapted from the maturity measurement scale set out in ISO 37004 (Governance Maturity Model – Guidance). It reflects the structured progression of the ISO standards, employing a maturity scale that ranges from Level 0 (No evidence) to Level 5 (Continual optimisation or improvement). It is grounded in the principle of evidence-based assessment, evaluating maturity through the lens of activities and inputs (Levels 1–2) that lead to measurable outcomes and impact (Levels 3–5). The application of this instrument in the first iteration of assessing AI maturity in South Africa indicated that the country is in the early stages of AI maturity, with seven of the eight domains assessed at Levels 1–2 on the maturity scale, and only one domain reaching Level 3. This suggests that all domains demonstrate some level of foundational ‘input activities’, signalling a positive intent and growing awareness of the importance and relevance of AI, with some initial movement towards demonstrating positive outcomes. Overall, the generally low maturity scores indicate that progress across the domains is fragmented. This could be attributed to still limited strategic ecosystem coordination, targeted investment and capability building. The evidence-based approach enables a reflection of the South African landscape and supports strategic alignment across government, industry, and academia. The framework and interactive dashboard – developed through this project to enable accessible visualisation of the data – serve as foundational reference points for supporting South Africa’s evolving AI ecosystem. An AI agent developed as part of the initiative facilitates regular updates, helping ensure that the AI MAF remains a relevant strategic tool." (Executive summary, pages 8-10) more

Download (11)

LAO P.D.R. Artificial Intelligence Readiness Assessment Report

Paris: UNESCO (2025), 72 pp.

"• In the legal context, Lao PDR’s regulatory framework for AI is still nascent. While the government is drafting a National AI Strategy, there are currently no laws addressing core AI issues such as transparency, accountability, or algorithmic redress. Existing frameworks like the Cybercrime La ... w, Data Protection Law and the Digital Government Decree cover general data and computing use but lack AI-specific provisions. AI is referenced in recent constitutional amendments and draft digital transformation policies, but there’s room for improvement in transforming those acknowledgements to clear legal mechanisms to protect citizens and bolster innovation. • For the sociocultural domain, despite progress in gender inclusivity, rural communities in Laos still face limited access to digital technologies due to infrastructure gaps, affordability, and low digital literacy. Successful collaborations with international organizations highlight what’s possible but also stress the need for ongoing investment in local capacity, outreach, and trust-building. Digital initiatives in health and efforts addressing concerns about technology’s environmental impact are evident, but they have not been specifically focused on AI. Digital transformation must also prioritize cultural preservation, including the development of culturally-aware datasets and NLP tools. Currently, most Lao language datasets and NLP tools are lagging participation from native Lao researchers, risking linguistic sovereignty and underrepresenting local dialects, norms, and values. • AI talent and innovation development in Laos needs strong support from its education and research ecosystems. University departments like Computer Science and Software Engineering are small and often siloed, with little crossdisciplinary collaboration. Research output needs some improvement. Promising developments in AI/ICT curricula in K-12 level are welcome, although their effectiveness must be closely monitored. • Laos’s digital economy remains small, constrained by a large informal sector, limited digital infrastructure, and low ICT investment, factors that hinder AI adoption and innovation. While some private companies in sectors such as telecom, banking, logistics, and manufacturing are beginning to use AI tools, and urban populations are gradually adopting app-based lifestyles, broader progress is limited by lagging human capital pipelines, minimal domestic investment, and an underdeveloped startup ecosystem. Foreign-developed platforms still continue to dominate the market, which if continued can limit local developers from developing capabilities and use cases. Aside from that, most Foreign Development Investment (FDI) remains concentrated in resource extraction, and budget shortfalls undermine the sustainability of e-government initiatives. • There is a critical need for a strong commitment to funding both the hard and soft infrastructure that supports AI. This includes enhancing cybersecurity capacity, reliable networks, and data centers and ensuring affordable access to computing infrastructure for AI practitioners and stakeholders." (Executive summary) more

External Link (33)

Índice Latinoamericano de Inteligencia Artificial (ILIA) 2025
Deep Insights

Álvaro Soto, Rodrigo Durán, Antonia Moreno, Sebastián Adasme, Sebastián Rovira, Valeria Jordán, Laura Poveda (coord.)

Santiago de Chile: Comisión Económica para América Latina y el Caribe (CEPAL); Centro Nacional de Inteligencia Artificial (2025), 195 pp.

"El Índice Latinoamericano de Inteligencia Artificial (ILIA) constituye un esfuerzo pionero para medir de manera sistemática el avance de la IA en 19 países de la región. El ILIA busca no solo monitorear el grado de preparación, adopción y gobernanza, sino también difundir buenas prácticas, ... además de ser un instrumento para orientar políticas públicas que incorporen la IA al modelo de desarrollo de los países de manera integral. El índice se organiza en tres dimensiones: Factores Habilitantes (infraestructura digital, talento humano y datos), Investigación, Desarrollo y Adopción, y Gobernanza. A partir de más de cien subindicadores, el ILIA 2025 ofrece un panorama detallado sobre avances, brechas y oportunidades estratégicas en el despliegue de la IA en América Latina y el Caribe. [...] Once hallazgos principales del ILIA 2025: i) El despertar de los adoptantes tardíos. Ecuador, Costa Rica, República Dominicana y Guatemala muestran mejoras aceleradas en conectividad, talento y estrategias de IA, reduciendo distancia con los países líderes, y abriendo oportunidades que parecían improbables para una cooperación regional. ii) Código abierto: la oportunidad para América Latina y el Caribe. El alto desempeño de Honduras, El Salvador y Cuba en producción de software abierto confirma que este modelo ofrece una vía estratégica para desarrollar soluciones locales, fomentar la colaboración regional y el aprendizaje colectivo, además de promover la transparencia algorítmica. iii) Muchos datos, pero poca disponibilidad. La región produce grandes volúmenes de información, pero su apertura y estandarización siguen siendo limitadas. Esto restringe la ciencia abierta, reduce la transparencia y frena el desarrollo de soluciones locales de IA. iv) Talento humano: alfabetización, poca especialización. La alfabetización en IA duplica a la formación profesional y cuadruplica al talento especializado. Aunque se amplían programas escolares y posgrados, persiste un embudo en la formación avanzada que limita el potencial de la región para producir soluciones propias. v) Talento e infraestructura para la soberanía digital. Brasil concentra más del 90% de la capacidad de cómputo de alto rendimiento de la región. Aunque Uruguay, Costa Rica y Colombia muestran mejores indicadores per cápita en Unidades de Procesamiento Gráfico (GPU por sus siglas en inglés), más de la mitad de los países carecen de infraestructura crítica, lo que refuerza asimetrías. Además, 13 de los 19 países no generan habilidades tempranas en IA en el currículo educativo y 11 no cuentan con programas de doctorado en IA en sus universidades. Sin fortalecer estas capacidades endógenas, la región corre el riesgo de generar soluciones de IA sesgadas y con baja pertinencia local. vi) IA Generativa: una oportunidad de democratización del acceso. La IA generativa ha democratizado el acceso y acelerado la adopción gracias a sus bajas barreras de entrada y a interfaces simples que permiten a no expertos usar modelos avanzados. Países como Chile, Costa Rica, Perú, Uruguay, Panamá y República Dominicana destacan en uso, y ALC es ya la tercera región del mundo en descargas de aplicaciones de IA Gen (15%–20%). Esta accesibilidad actúa como catalizador y, junto con el desarrollo de software y el código abierto, compensa déficits de cómputo, abriendo oportunidades para toda la región. vii) IA y participación ciudadana: un cambio de paradigma desaprovechado. Aunque la IA podría revolucionar la democracia digital, su uso se limita en la mayoría de los países a chatbots informativos y no en herramientas de consulta, de rendición de cuentas o de co-creación de políticas públicas. Colombia, México y Perú muestran algunos avances, pero ocho países no registran ningún caso de aplicación. viii) Investigación escasa y concentrada. Brasil y México reúnen el 68% de los investigadores activos en IA y, junto a Colombia, Chile y Argentina, concentran el 90% de las publicaciones. Solo siete países participan en conferencias internacionales de élite, con predominio de Chile y Brasil. La escasa presencia en foros académicos de alto impacto reduce la visibilidad y la transferencia de conocimiento de la IA de la región; la actividad académica sigue rezagada y requiere mayor atención en la mayoría de los países. ix) Gobernanza en IA: mucho plan y poca acción. Nueve países cuentan con estrategias nacionales de IA, pero solo unos pocos han asignado presupuesto, definido planes de implementación o establecidos indicadores de impacto. Sin mecanismos de ejecución, estas estrategias corren el riesgo de quedarse en meras declaraciones. Al mismo tiempo, mientras a nivel global avanza la apertura e interoperabilidad de datos, la región mantiene un ritmo lento, centrado en debates normativos y éticos. El verdadero desafío es equilibrar la regulación con políticas orientadas al desarrollo y a la implementación efectiva de la IA, de modo que esta tecnología genere valor sin frenar la innovación. x) Sustentabilidad: un llamado urgente. Solo cuatro países (Brasil, Chile, Colombia y México) muestran industrias de centros de datos robustas, y uno de cada cinco centros cumple con estándares internacionales de sustentabilidad. La mayoría de las políticas de IA y digitalización omite el impacto ambiental, lo que puede entrar en contradicción con los objetivos de sostenibilidad medioambiental de la región. xi) Emprendimiento en IA: una oportunidad para la región. Los 19 países ILIA concentran sólo el 1,12% de la inversión global en IA y cuenta con empresas unicornios en apenas seis países. No obstante, el ILIA muestra que no es necesaria una matriz productiva compleja para fomentar el desarrollo de empresas de IA locales y el uso de esta tecnología en el sector productivo. Entornos favorables de innovación y emprendimiento, inversión y políticas de desarrollo productivo pueden detonar dinámicas innovadoras." (Resumen ejecutivo, páginas 15-16) more

External Link (34)

Mentorship toolkit: Creating sustainable business and funding models with open source AI

Villgro Africa

Bonn; Eschborn: Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ); FAIR Forward - Artificial Intelligence for All (2024), 48 pp.

"This toolkit is designed to help you create and run a successful mentorship program that will support and empower opensource AI entrepreneurs to develop sustainable businesses and funding models for their projects. Open source AI is a rapidly growing field that offers many opportunities for innovat ... more