Diplomacy & Defense Think Tank News

The Turkish government’s declaration of 2025 as the “Year of the Family” reflects an increasingly authoritarian approach to managing demographic decline amid conditions of economic strain. Falling fertility rates are framed as a strategic challenge to national resilience, giving rise to a policy agenda that positions the family as the cornerstone of demographic recovery. At the same time, anti-gender politics has moved beyond ideological rhetoric towards a broader governance strategy linking demographic management, fiscal constraints, and authoritarian consolidation. For Germany and the European Union (EU), understanding these developments is essential for engaging Turkey on democratic governance, social policy, and gender equality.

Ahmed al-Scharaa ist Islamist, will aber keine Theokratie, sagt die Wissenschaftlerin Muriel Asseburg. Die Lage für die Kurden habe sich durch das Abkommen vom 30. Januar gebessert. Vor Abschiebungen nach Syrien warnt sie.

Der Krieg gegen den Iran zerreißt die westlichen Gesellschaften. Er ist völkerrechtlich kaum zu rechtfertigen, doch sein Abbruch wäre das schlimmste aller Ergebnisse. , Der Krieg Israels und der USA gegen den Iran spaltet die westlichen Demokratien wie kaum ein geopolitisches Ereignis der jüngeren Geschichte. Die Frage, ob dieser Krieg gerechtfertigt ist, lässt sich nicht mit einem einfachen Ja oder Nein beantworten. Wohl aber lässt sich sagen, was das Schlechteste ...

Die Bilanz von vier Wochen Krieg fällt für die iranische Bevölkerung verheerend aus, sagt die Iran-Expertin Azadeh Zamirirad im Interview. Was bedeutet das für den Wunsch nach einem Regimewechsel und für die Opposition?

Ce sont des révélations faites par un consortium de médias internationaux, dont les Observateurs de France 24. Un réseau d'agents russes spécialisés dans la désinformation propage sa propagande anti-Ukraine dans toute l'Afrique et en particulier en Côte d'Ivoire depuis plusieurs années. Son nom : la "Compagnie". Décryptage de Jules BOITEAU dans Info Intox. 

Seit Russland 2022 seinen vollumfänglichen Aggressionskrieg gegen die Ukraine begonnen hat, verfolgt die EU ihre Erweiterungspolitik unter geopolitischen Vorzeichen. Wie es bei der Europäischen Kommission und bei Kandidatenstaaten inzwischen heißt, könnte die EU ab 2028 neue Mitglieder aufnehmen; die Ukraine fordert sogar, 2027 beizutreten. Doch hohe Hürden stehen der Erweiterung im Weg. In der EU stagniert der Reformprozess, der ihre Handlungsfähigkeit verbessern und sie für eine Mitgliederzahl von 30 oder mehr fit machen soll. Beitrittsreife der Bewerber und Absorptionskraft der Union sind aber Schlüsselkriterien, die die deutsche Europa-Politik traditionell hochhält. In der EU wird nun intensiv über schrittweise Integra­tion, neue Schnellspuren und zusätzliche Sicherheitsklauseln debattiert, um den Ziel­konflikt zwischen geopolitischer Dringlichkeit und integrationspolitischer Solidität zu minimieren. Gegenüber den Staaten des Westbalkans sollte die EU am erprobten Beitrittsprozess festhalten. Im Fall der Ukraine ist die Lage so akut, dass es die poli­tische Verpflichtung, das Land aufzunehmen, zu bekräftigen gilt. Als Vorstufe zur Mitgliedschaft sollte die EU Kyjiw eine Beitrittsassoziierung neuen Typs anbieten, die auch eine Sicherheits- und Verteidigungspartnerschaft umfasst.

La Libye marque cette année les 15 ans depuis sa révolution qui a déposé Mouammar Kadhafi. Malgré une relative stabilité sécuritaire, le pays reste divisé en deux. L'ouest est dirigé par un gouvernement reconnu par la communauté internationale. L'est et une grande partie du sud sont, eux, dirigés par le régime militaire du maréchal Khalifa Haftar. Reportage à Benghazi, dans l'est du pays.

The Pax Americana, which guaranteed the security of Germany and Europe after the Second World War, is coming to an end. Europe can no longer rely on its alliance and partnership with the United States. Even before Donald Trump’s second presidency, Europe was seeking to reduce its dependency on Washington for peace, democracy and prosperity. This cannot be achieved overnight, and will require a significant increase in material resources and strategic thinking over the next five to ten years. The fourteen contributions to this SWP Research Paper show how differ­ent the starting conditions for developing transition strategies are, depending on the policy area and challenges. Europe needs to consider all the options: with, without or even against Washington. The analyses of Europe’s agency and the scope for European policy towards Russia, the Middle East and China are very wide-ranging. Trump’s logic of quick deal-making and unilateralism under the banner of “MAGA” often collides fundamentally with the EU’s multilateral foreign and security policy, which is bound by international law, and its commitment to sus­tainable peace. The keywords for the urgent reorganisation of security in Europe are: Europeanisation of NATO, strengthening Europe’s own military capa­bilities, new leadership constellations for security policy in Europe, and resilient governance in technology and cybersecurity. Even at this geopolitical turning-point (“Zeitenwende”), the EU should continue to develop its soft power. When it comes to the crucial questions of global governance – from UN and international law to trade, climate and energy policy – Europe must find new partnerships and, if necessary, new institutional solutions without and against the United States.

Fier de la présence sur son territoire de nombreuses tortues marines, le Gabon s'engage dans la préservation des différentes espèces. Chaque jour, pendant la saison des pontes, des ONG et des écogardes de l'Agence nationale des parcs nationaux patrouillent sur les plages du pays pour protéger les œufs et mettre à l'abri les plus menacés.

Un réseau d’agents russes spécialisés dans les opérations de désinformation, la "Compagnie", a lancé une vaste campagne de propagande contre l’ambassade d’Ukraine en Côte d’Ivoire. Grâce à une fuite de documents inédits reçus par le média panafricain The Continent, la rédaction des Observateurs a pu enquêter, avec un consortium de médias, sur les rouages de cette organisation. 

Une fuite de documents confidentiels d'un réseau d'agents russes sur le continent africain révèle la manière dont la Russie a réussi à diffuser des centaines d'articles dans 35 médias francophones ouest-africains, parfois à leur insu. Avec un consortium international de médias coordonné par Forbidden Stories, la rédaction des Observateurs de France 24 a analysé ces documents stratégiques, qui montrent en miroir la porosité de ces médias face aux ingérences extérieures.

La France a retiré à l'Afrique du Sud son invitation au G7 de juin prochain, qui doit se tenir à Évian. Pretoria a d'abord dénoncé jeudi des pressions américaines pour expliquer cette mise à l'écart, avant que le président sud-africain lui-même fasse marche arrière. Washington, de son côté, nie toute implication. Paris assure n'avoir "cédé à aucune pression", mais avoir fait le choix d'un "G7 resserré".

Dans la région des Grands Lacs, Reporters sans frontières (RSF) sort un nouveau rapport intitulé "Dans la peau d'un journaliste des Grands Lacs". Nouvelle sonnette d'alarme pour dénoncer les violences auxquelles font face les journalistes dans cette région, la plus dangereuse du continent selon RSF. Six pays sont concernés et la RD Congo tient une place particulière.

The parliamentary elections held in Bangladesh on 12 February 2026 marked a turn­ing point in the country’s recent history. In August 2024, Prime Minister Sheikh Hasina, who had served for a total of 20 years, with one interruption, was removed from office. An interim government led by Nobel Peace Prize laureate Muhammad Yunus subsequently took power. Following two years of political uncertainty, it suc­ceeded in organising elections that were procedurally sound. Bangladesh’s political system has thus demonstrated a considerable degree of institutional resilience. At the same time, the election outcome raises new questions regarding the future of democratic pluralism in the country. The decisive election victory of the Bangladesh Nationalist Party (BNP) means that one of the country’s two historic political family dynasties will remain at the helm of government. The Muslim fundamentalist camp gained vote share, which could negatively impact the role of women in society as well as complicate Bangladesh’s renewed rapprochement with neighbouring India.

Bei der sogenannten Riester-Reform, über die am morgigen Freitag im Bundestag beraten wird, hat die SPD die Einführung eines staatlich organisierten Standardfonds für die private Altersvorsorge durchgesetzt. Peter Haan, Leiter der Abteilung Staat im Deutschen Institut für Wirtschaftsforschung (DIW Berlin), kommentiert diesen Vorschlag wie folgt: 

Viele Elemente des vorgeschlagenen Altersvorsorgedepots haben das Potenzial, die Verbreitung der privaten Altersvorsorge zu stärken. Positiv hervorzuheben ist die Einführung eines staatlichen Standarddepots – allerdings ist die Kostengrenze von einem Prozent immer noch ziemlich hoch. Auch der Wegfall der Beitragsgarantie ist grundsätzlich sinnvoll: Er ermöglicht renditestärkere Anlagestrategien und macht das Produkt langfristig attraktiver.

Allerdings löst diese Reform das wichtigste Problem nicht. Die Verbreitung der privaten Altersvorsorge ist heute sehr ungleich verteilt – und besonders gering bei Haushalten mit niedrigen Einkommen, also genau dort, wo das Risiko von Einschnitten im Lebensstandard im Alter und die Altersarmut am größten ist. Freiwillige Anreize haben daran seit zwei Jahrzehnten nichts geändert. Wer diese Lücke schließen will, kommt an einer Verpflichtung nicht vorbei – ob über die betriebliche oder die private Säule, darüber lässt sich streiten. 

Hinzu kommt ein Verfahrensproblem. Die Bundesregierung hat eine Alterssicherungskommission eingesetzt, die bis Mitte 2026 klären soll, wie viel Altersvorsorge insgesamt notwendig ist und wie die drei Säulen zusammenwirken sollen. Ein Gesetz zur dritten Säule jetzt zu beschließen, setzt Fakten, bevor die Gesamtarchitektur steht. Das ist keine gute Grundlage für eine Reform, die Jahrzehnte wirken soll.

Der Artikel „Public Communication and Collusion: New Screening Tools for Competition Authorities“ von Tomaso Duso, Carl Kreuzberg, Joseph Harrington und Geza Sapi hat den Concurrences Antitrust Writing Award 2026 in der Kategorie „Wissenschaftliche Artikel – Koordinierte Praktiken“ gewonnen. Die ...

Rapid advancements in the field of Artificial Intelligence (AI) are expected to deliver significant economic and productivity benefits. The EU, which risks returning to a low-growth path, would benefit from the development and adoption of AI systems. However, in almost all key metrics, the EU lags far behind the US and China in AI. To reverse this trend, Europeans must tackle the structural causes that have held back AI companies, including Single Market barriers and the lack of VC funding. Europeans should also focus their efforts on open-source AI leadership to boost business adoption. Greece would also benefit significantly from these EU-wide reforms. But at the same time, Greece will need to pursue its own vertical niche in AI in order to become an AI exporter.

• The capabilities of AI systems have improved significantly in the last decade. Today’s AI systems can process different types of information, such as text, video, audio, code, and solve complex reasoning challenges. 

• Due to these technical advances, AI systems can help businesses automate tasks, enhance human workers and deliver new services. This AI transformation could deliver a 7-15% increase in global GDP, in the next decade. 

• However, as the European Union (EU) lags behind the development and adoption of AI systems, it may lose out on the economic and strategic benefits that the technology provides. 

• The EU’s AI lag could be attributed to structural barriers that have prevented the creation and scaling of AI solutions. These include the barriers of the Digital Single Market, the lack of access to high-quality funding and the inability to translate Europe’s scientific excellence into concrete market outcomes. 

• The EU’s current approach to AI, which includes initiatives such as the AI Act and the AI Continent Action Plan, must be updated accordingly to address these structural challenges and provide effective incentives for the emergence of European AI frontrunners. 

• As part of the EU, Greece would benefit significantly from such reforms. At the same time, Greece’s limited resources mean that the country’s position in this field can only be improved by pursuing advantages in vertical AI domains, such as agriculture and climate resilience.

Read here in pdf the Policy paper by Giorgos Verdi, Policy Fellow at European Council on Foreign Relations (ECFR).

Technical advances in AI

The capabilities of AI systems have been progressing rapidly over the last decade. In 2017, AI exceeded human performance in both image and speech recognition. By 2022, the release of the AI system ChatGPT had captured the world’s attention with its ability to perform language-based tasks. The virality of ChatGPT’s release also showcased to the wider public that AI systems had achieved the capacity to perform some economically-valuable tasks.

Since then, AI development has been marked by further qualitative leaps. Today, cutting-edge AI systems are multimodal: they can process text, audio, images, video and computer code. They also showcase notable improvements in their ability to solve complex reasoning challenges by breaking down multi-step problems into intermediate steps. Leveraging these algorithmic innovations, AI systems achieved gold-medal performance in the International Mathematics Olympiad in 2025 and have surpassed PhD-level human experts in science tests.

Besides algorithmic innovations, improvements in AI capabilities have also been made due to the growth in computational resources and available training data. This correlation between more compute, data, parameters and model performance has been described by researchers as ‘scaling laws’. Under the current paradigm, developers are in need of more powerful computer resources and new data sets in order to train ever larger models and achieve new breakthroughs.

Despite these trends, AI development is still faced with certain limitations. AI systems often generate misleading outputs and fabricated data. The spatial capabilities of AI – the prerequisite for the execution of physical tasks – remain far from human level. AI systems also perform poorly in long-horizon planning and execution. This paradox has led researchers to characterise the status of the AI frontier as uneven or ‘jagged’. AI excels in domains where machine-readable data is available but underperforms in tasks that are unstructured or ambiguous.

Economic implications of AI

Technical advancements in AI systems can have a significant economic impact, both in the world and in Europe. AI can provide economic benefits by helping businesses to automate tasks, enhance human workers’ capabilities, and roll out new services. As a result, AI’s economic impact is expected to span multiple domains and industries. The OECD predicts that AI can be used in most industrial activities, from optimising machine systems to improving industrial research and development (R&D). AI could also boost automation in manufacturing, ensure timely and cost-effective maintenance, and guaranteeing stronger quality of products and processes. In healthcare, AI is already helping doctors to scan medical images in order to deliver faster and more accurate diagnosis. AI is also revolutionising drug discovery by extracting novel insights from complex biomedical data. In automotive, AI is used in driverless cars to recognise traffic signs and pedestrians.

AI-driven transformation across industries is then expected to drive economic growth. However, the magnitude of such growth has been contested. On one hand, Goldman Sachs predicts that AI could drive a 7% increase in global GDP (or almost $7 trillion) and lift productivity growth in the United States by 1.5% annually in the next decade. In a similar line, the McKinsey Global Institute predicts AI to increase global GDP by 1.2% annually and deliver an economic output of $13 trillion in the next five years. PwC’s estimates are even higher, with an expected increase to global GDP by 8-15%, in the next 10 years.

Other reports predict that AI will have a more modest macroeconomic impact. MIT Professor Daron Acemoglu estimates that AI will not lift global GDP more than 1-2% over the next decade with less than 0.1% annual gain in productivity. However, it must be noted that Acemoglu’s estimates rest on the assumption that only 5% of tasks can be performed more cheaply by AI systems than human workers. In turn, this estimate was based on a 2023 study when AI systems were much less capable. Finally, other approaches have tried to reconcile these differences and consider a scenario where AI systems raise US productivity by 0.5%. Under this assumption, real GDP per-capita could nearly be $7000 higher by 2035.

The uncertain impact of AI on economic growth can also be attributed to the ‘productivity paradox’ ingrained to technological transformations. For example, the productivity gains of electricity took 40 years to fully materialise after the first generating station was built by Thomas Edison. Similarly, information and communication technologies didn’t lead to expected productivity gains in the 1970s and 1980s. As Robert Solow described this productivity paradox in 1987: “You can see the computer age everywhere but in the productivity statistics”.

In order to explain why the productivity paradox occurs, researchers have argued that transformative technologies follow a ‘J-curve’ trajectory. The extensive investments which are required to integrate transformative technologies into organisations are often underestimated. Historically, there have been periods of considerable length in which unmeasured inputs are being built in order to complement the new transformative technology. For example, AI systems that are used to boost productivity also require investments in staff training, workflow redesign, data and compute infrastructure. These bottlenecks are likely to cause an initial dip in productivity – the downward slope of the J-curve. However, once these adjustments are made, firms tend to experience a productivity upswing.

If then, AI systems are to drive economic growth – albeit within a longer timeline – it could present a significant economic opportunity for the European Union (EU).

If then, AI systems are to drive economic growth – albeit within a longer timeline – it could present a significant economic opportunity for the European Union (EU). Since 2000, productivity in the Eurozone has been on average 0,5% lower each year than the United States (US). This is mainly due to productivity growth in the information and communication sector and in professional services being weaker in the EU. As these two sectors make strong use of technological innovations, this shows that the EU has been less able to make use of digital technologies. If this trend persists, the consequences could be dire for the EU’s prosperity. However, the development and adoption of AI systems in the EU could help reignite productivity growth.

Besides productivity, AI could also contribute to address several of the EU’s challenges. The EU is faced with a demographic challenge, as the populations of several EU Member States peaked prior to 2025 and have now started to fall. The European population is also ageing, creating generational imbalances. This trend will generate fiscal challenges, through pressures on pension spending and the effects on the tax base. AI systems, and their potential productivity benefits, can help offset some of the pressures that demographic trends will impose on the EU’s economy.

In addition, the EU has launched an agenda of military rearmament in the face of Russia’s increased aggression and the gradual retreat of the US from European security. To achieve this, the EU adopted ReArm Europe which aims to mobilise €800 billion of defence expenditures. AI systems can play a role in this endeavour as their integration within military forces promises to provide cost-effective solutions. A report by RAND, the American think-tank, estimates that AI solutions can save US Air Forces $25 million per month – only on the maintenance of A-10C aircrafts. Similarly, the integration of AI into French Ceasar artillery pieces could provide a 30% savings in ammunition. Overall, the military applications of AI systems are likely to extend the economic benefits of the technology beyond the civilian realm. As a result, AI systems can also offset some of the fiscal pressures that increased military spending presents to European governments.

All in all, the technical advances in AI systems promise to transform industries and provide productivity benefits. Even if the optimistic projections of experts are not realised, the EU would benefit significantly from the development and adoption of AI systems. The solutions to the EU’s major challenges – from its growth and demographic problems to rearmament – should take into consideration the novel and advanced capabilities of AI systems.

The EU’s challenges in AI

Despite the potential benefits of AI systems, the EU is faced with a series of challenges in their development and adoption. As a result, the EU and its Member States may end up missing on the advantages that the technology could provide.

…the EU lags behind both the United States and China in most metrics of AI development and innovation.

Firstly, the EU lags behind both the United States and China in most metrics of AI development and innovation. According to Stanford’s AI Index, the US developed 40 cutting-edge AI models in 2024. In the same year, China developed 15 while the EU developed 3 – all of them by the French company Mistral. In the same year, US AI companies attracted $109 billion in private investments, while EU AI companies attracted $14.9 billion. The EU also lags behind the US in terms of AI entrepreneurship: 1143 AI companies were newly funded in the US in 2024, compared to 335 in the EU.

China has also surpassed the EU in terms of scientific publications in AI with 23,695 related publications in 2024, compared to 10,055 papers published by EU scientists. In addition, China accounts for 69.7% of the world’s AI patents while the country also dominates in AI robotics, surpassing the rest of the world combined in installations. According to an MIT study, Chinese has also overtaken both American and European open-source AI models, in terms of worldwide popularity.

As a result of their distinctive advantages, the US and China are home to multiple companies that develop frontier AI models. The US boasts the presence of companies such as OpenAI, Google, Anthropic, xAI and Meta. Meanwhile, Chinese start-ups DeepSeek and Moonshot have released models that are on par with those of OpenAI, while companies Tencent and Alibaba have acquired world class capabilities.

…only one European company currently competes at the frontier of AI development: the French start-up Mistral.

As a result of the EU’s gaps, only one European company currently competes at the frontier of AI development: the French start-up Mistral. However, Mistral’s models have not yet matched or surpassed the performance of their American and Chinese counterparts. In addition, Mistral’s limited funding has put the survival of the company in question. While OpenAI’s market valuation reached $500 billion in October 2025, Mistral has reached a market valuation of $14 billion. As a result, there have been reports that US company Apple has explored the acquisition of Mistral. If such a deal was to be realised, then the EU would be left with no homegrown AI developers at the frontier.

Besides AI models themselves, the EU lacks capacity in various parts of the AI value chain. AI chips that are used to train AI models are designed by US companies Nvidia and AMD and manufactured in Taiwan by TSMC. AI data centres are also increasingly concentrated in the US which hosts 75% of the world’s AI supercomputers, according to some sources. The EU, on the other hand, hosts 2-5% of the world’s AI supercomputers.

Finally, the EU has fallen short of its objectives when it comes to AI adoption. Even when the EU imports AI models and capabilities from elsewhere, businesses seem reluctant to take advantage of its productivity benefits. According to the EU’s Digital Decade report, 18% of European small and medium-sized enterprises (SMEs) have taken up AI until 2025. However, the EU’s objective is for 75% of SMEs to take up AI until 2030. It is highly unlikely that the EU will reach its target. In the US, different sources have estimated SME AI adoption to be between 20% and 40%, due to methodological differences.

Slow adoption of AI means that the productivity dividends over the next few years could be lost.

These challenges in the development and adoption of AI will hinder the ability of the EU to take advantage of the economic benefits of AI. Slow adoption of AI means that the productivity dividends over the next few years could be lost. Lack of AI development capabilities also means that Europeans will have little control over the direction of the technology. In 2025, the Biden administration decided to limit the export of American AI chips to 18 EU Member States, including Greece. Practically, these restrictions would have meant that the majority of EU countries would be unable to acquire AI data centres. Although the Trump administration revoked President Biden’s decision, this incident showcased how a lack of indigenous AI development capabilities can leave European vulnerable to external shocks.

Lastly, the above metrics point towards a growing gap between the US and China on one hand, and the EU on the other. This means that the relative economic and military strength of Europe is likely to be diminished as the two superpowers take advantage of AI. Moreover, other middle powers, such as Saudi Arabia and the United Arab Emirates have made significant efforts to become more competitive in the technology and direct vast amounts of capital into AI capabilities. This trend is likely to put more pressure to the EU who finds itself competing with an ever growing amount of actors in the domain of AI.

The causes of EU’s AI lag

Even if the EU falls behind China in terms of AI publications, the bloc remains a world-leading centre for AI research, talent and universities. As a result, the EU’s AI lag is caused by its inability to translate scientific excellence into concrete market outcomes. In order to improve the EU’s position in AI, the structural reasons that have led to this situation must be understood.

One of the most common points of criticism against the EU’s digital policy has been its focus on imposing strict regulations and governance frameworks on innovators. In the 2019-2024 mandate, the European Commission adopted 93 pieces of legislation that were of relevance to technology innovators. According to Mario Draghi, these Regulations and Directives have been restrictive and inconsistent, and in turn, have hindered innovative companies in Europe at every stage of their growth.

Some experts have tried explaining the EU’s AI lag based on this argument. In 2024, the EU adopted the Artificial Intelligence Act, the world’s first law to regulate AI systems. The regulation seeks to protect consumers and citizens from the harms that high-risk AI systems impose. However, it could also impose high costs of compliance and may have driven AI innovators away from building models in the EU. According to some estimations, the costs of complying with the AI Act’s legal obligations for so-called ‘high-risk’ AI systems may reach the range of €193,000 and €330,000. Besides the AI Act, innovators are also faced with legal uncertainty stemming from other regulations.

…effective regulation can provide benefits for European AI businesses.

However, the EU’s regulatory approach to AI can also bring benefits to businesses and consumers. Firstly, adherence to strict legal standards can help boost consumer trust in AI. But if left unchecked, the risks that stem from the technology can deter its adoption. Thus, effective regulation can provide benefits for European AI businesses. Secondly, the EU’s initiative to legislate on AI harmonised rules across Member States and prevented the risk of regulatory fragmentation. Thus, the alternative to the EU’s AI Act would have likely been the creation of 27 different regulatory regimes, which would have imposed a higher burden on businesses.

In addition, the EU’s innovation lag cannot be attributed to regulation alone. Before 2016, the EU’s Single Market was not heavily regulated and was only governed by two pieces of legislation: the 2000 e-Commerce Directive and the 1995 Data Protection Directive. Despite the lack of digital regulations, the EU still failed to produce successful technology companies in that period. Most of the EU regulations commentators have criticised, such as the AI Act and the Digital Services Act (DSA) were adopted in the early 2020s. As a result, other structural causes should be examined to understand the EU’s inability to nurture frontrunners in AI and in digital technologies more broadly.

Columbia Professor Anu Bradford has argued that alternative factors, other than regulation, have held the EU’s digital and AI economy back. One of the most major impediments has been the lack of a European Digital Single Market. While some steps have been taken towards further integration, the European market remains fragmented for AI companies that require scale in order to grow. Companies are required to navigate 27 different legal regimes for businesses, including differences in labour laws and tax systems. According to the IMF, Europe’s internal barriers are equivalent to a tariff of 110% for services. That puts European start-ups in a disadvantage compared to their US and Chinese counterparts that begin their growth journeys in largely homogenous markets.

Another significant impediment has been Europe’s shallow and fragmented capital markets.

Another significant impediment has been Europe’s shallow and fragmented capital markets. Innovative companies in the US have benefited largely from the existence of a vibrant Venture Capital (VC) ecosystem, in which firms are more willing to invest in high-risk, high-reward technology start-ups. Whereas the US represents 52% of the world’s venture capital market, the EU only represents 5%. European companies also struggle to attract American VC investments, as geographical proximity plays a role in firms’ decisions, according to empirical data. As a result, European innovative companies have to rely on risk-averse banks in order to finance their operations.

Taken together, these alternatives variables provide a more comprehensive explanation of the EU’s AI lag. While the costs of complying to the EU’s digital regulation can be excessive, these laws also provide trust for consumers and certainty for businesses. At the same time, the EU’s internal fragmentation and the lack of VC funding are more likely to be responsible for the lack of European AI capabilities. Any effort to reverse Europe’s AI gaps must start by addressing these structural challenges.

The EU’s AI strategy and the way forward

Europe’s AI lag has prompted officials in Brussels to expand the EU’s AI strategy beyond product regulations and adopt new sets of measures to spur innovation. In April 2025, the European Commission adopted the AI Continent Action Plan to accelerate efforts across five domains: compute infrastructure, data, skills, and compliance. While the AI Continent Action Plan is a welcome step, it does not effectively address the root causes of the EU’s bottlenecks, as described above. Therefore, an additional set of policy measures will be required to improve the current approach and allow Europeans to reap the strategic and economic benefits of AI.

In the first quarter of 2026, the European Commission is expected to put forward a new proposal that will aim to tackle internal barriers for start-ups and lead to a more integrated Digital Single Market. This has been dubbed the ‘28th regime’ as it will establish a single and harmonised set of EU-wide rules that would allow innovative AI companies to scale, instead of having to deal with 27 different regimes. At the 2026 World Economic Forum, the President of the European Commission added that the proposal would allow anyone to register a company fully online and within 48 hours.

While these are steps in the right direction, it is not clear yet whether the initiative will take place via a Regulation or through a Directive. The European Parliament’s opinion to introduce a Directive is likely to lead to a fragmented transposition and kill the spirit of the policy. It is then important that the European Commission swiftly implements this idea as a Regulation.

For AI companies to scale across Europe’s 27 different markets, it is also urgent that they are able to raise private funding. The American VC ecosystem has benefited from leveraging institutional investments, such as those coming from pension funds and insurance funds. This model is challenging for the EU to replicate in the short-term as its pool of pensions is smaller. Member States could this aim to deliver attractive returns to pensioners and finance innovation by encouraging supplementary pension schemes through beneficial tax treatments or applying opt-out mechanisms.

However, Europe still benefits from a deep pool of insurance capital, which remains underexposed to VC. Member States should signal their political backing to encourage insurance funds and other institutional investors to rebalance their portfolios towards innovation. A mere 0.1% shift in asset allocation of insurers assets would channel an additional €10 billion into Europe’s VC ecosystem.

Besides market access and investments, innovators will benefit from regulatory certainty. Although the EU’s AI Act should not be held responsible for the bloc’s AI lag, it could still prevent the creation of new AI firms if its implementation is not done well. To this end, the European Commission introduced the Digital Omnibus on AI Regulation with the aim to simplify the AI Act and ease the burdens on innovators. However, this proposal has come under scrutiny from civil society organisations as it threatens to undermine critical safeguards, such as the protection of personal data. While the EU should pursue simplification, especially in compliance reporting requirements for SMEs, the removal of the AI Act’s safeguards could hurt the already declining consumer trust in AI systems.

But even if the above measures spur the creation of new AI firms, these innovators are expected to face some additional challenges. The EU’s limited compute capacity in AI, as described above, means that these companies will be forced to form a partnership with an American large company in order to access compute resource and train their models. In order to avoid these dependencies form consolidating, it then becomes urgent that the EU be able to offer its own compute resources to its aspiring frontrunners.

Here, the AI Continent Action Plan addresses this gap with the EU’s AI Gigafactories and AI Factories programme . The Gigafactories initiative specifically will receive €20 billion investments in order for each to contain 100,000 cutting-edge chips and help train large and complex AI models. However, industry representatives have expressed concerns that no more than 70 companies or consortia will be able to take advantage of such infrastructure. Therefore, it is crucial that these efforts are complemented by a concrete set of incentives for the creation of new AI start-ups and firms.

…there is room for European support to fund new breakthroughs and link the development of AI models to existing demand.

It then remains to be seen what sort of models European start-ups will build on top of this infrastructure. While the EU and its institutions should refrain from choosing champions, there is room for European support to fund new breakthroughs and link the development of AI models to existing demand.

In November 2025, France and Germany announced an upcoming Frontier AI Initiative that will be launched in Q1 2026, in collaboration with the European Commission. According to reports, it will create a nonprofit lab that will drive research which has either been overlooked or is not viable for market players to pursue. If done effectively, this work can be crucial as more and more experts have argued that the current paradigm of AI development is reaching its limits. The Frontier AI Initiative should pursue strategic bets on frontier AI, including bets on spatial and physical intelligence.

Finally, the EU should pursue better links between AI development and AI adoption in order to ensure that the technology is viable and that it yields concrete benefits for Europeans. In October 2025, the European Commission published the Apply AI Strategy which included sectoral flagship initiatives that aim to boost AI adoption across 10 industrial sectors. Moreover, the Apply AI Strategy promotes a ‘buy European’ approach for the public sector with a focus on open-source solutions. This last part is likely the most crucial component of the Strategy and should be extended to the private sector as well for the EU to fix its AI adoption gap.

The EU should set specific quotas for its AI funding instruments – whether the AI Gigafactories or investAI – to be dedicated to the development of open AI solutions.

Open-source AI systems and open weights AI systems grant users access to study how the system works and modify it for any purposes. As a result, they are highly customisable and cheaper that proprietary AI modes. An MIT study found that closed AI models cost users six times as much as open ones despite offering only moderate performance advantages. As a result, a proliferation of open AI models is likely to boost adoption within Europe’s low-resource actors, such as SMEs. In order to promote the development of European-made open AI models, and mitigate the surging popularity of Chinese open models, there is a need for concrete set of incentives that will encourage start-ups to follow this path. The EU should set specific quotas for its AI funding instruments – whether the AI Gigafactories or investAI – to be dedicated to the development of open AI solutions.

In conclusion, the way ahead for the EU’s AI strategy lies in breaking away from the false dilemma between regulation and accelerated innovation. The competitive dynamics between the US and China on AI have prompted some experts and industry representatives to push back against regulatory safeguards. More recently, the feud between the US Department of War and the AI firm Anthropic exemplified this debate, as the firm refused to drop its safety red lines over the use of its tools by the US military.

But as seen above, effective guardrails can boost trust and propel innovation forward. In contrast, an unregulated environment is likely to generate mistrust and public backlash. Five times as many Americans are concerned than are excited about AI, which has led to calls to slow down AI acceleration in the US. Even in the realm of defence, where safety red lines can be perceived as inherent weaknesses, the absence of guardrails can manifest as inefficient resource allocation, friendly fire incidents or the misidentification of targets. The EU should continue to pursue its trustworthy approach to AI development at the same time as new policies are implemented to boost innovation within that framework.

Greece’s AI challenges and prospects

As an EU Member State, Greece is facing many of the challenges mentioned above. As a smaller economy with a limited market of business and consumer customers, Greece would benefit from these broader reforms at the EU level. For example, while Greece is home to an impressive number of 188 AI start-ups, most of them are concentrated in early stages of funding. This trend aligns with the structural challenge that Europe faces in attracting late-stage investment and turning start-ups to scale-ups. As a result, Greece should promote relevant reforms at the EU-level. The upcoming Greek Presidency of the Council of the European Union in the second half of 2027 present a window of opportunity for the country to set the agenda on these points.

Besides EU-level reforms, much can be done at the national level in order for Greece to seize the opportunities of the AI revolution. It is noteworthy that at the 2023 NeurIPS – the world’s top AI conference – 11% of the top papers selected for an oral presentation had a Greek co-author. The selectiveness of NeurIPS indicates that Greece produces world-class AI researchers and developers. However, it is also the case that Greek AI talents often work and conduct research abroad. The main aim of Greece’s AI strategy should then be to become attractive to its own AI professionals and provide them with the necessary conditions that will turn Greece into an AI exporter.

The resource-intensive nature of AI development means that smaller countries can only improve their position through dominance of specific vertical AI industries. 

First of all, this would require that Greece finds and pursues its own distinct niche in AI. The resource-intensive nature of AI development means that smaller countries can only improve their position through dominance of specific vertical AI industries. To do this best, smaller countries need to use their national context to their comparative advantage. In 2024, the High Level Committee on AI published a Blueprint for Greece’s AI Transformation. While the document explores how Greece can enhance the business ecosystem, strengthen education, and support public administration, it refrains from defining and pursuing distinct advantages in niche areas of AI development within the private sector.

…as a significant agricultural producer within the EU, Greece could focus its limited resources on developing and exporting AI models in agriculture.

For example, as a significant agricultural producer within the EU, Greece could focus its limited resources on developing and exporting AI models in agriculture. This would empower Greek farmers through precision farming techniques and AI robots that can perform field work. It would also provide added value to Greek AI developers who would be able to collect high-quality data from the field and improve AI services in agriculture both in Greece and abroad. Initial support from the Greek Government and the EU may be required to help connect farmers to AI solutions and compensate for the lack of skills and trust. But as such solutions begin to scale, this market gap will likely be filled by the AI developers themselves who could become able to build AI solutions on-site and in close coordination with farmers.

Another example of a distinct niche that Greece could pursue is AI for climate resilience. AI systems can improve environmental monitoring and enhance early-warning systems for fires, heatwaves, draughts and water management. The development of Greek AI solutions that help predict such events can enhance preparedness and improve crisis relief for local communities. Moreover, Greek AI developers would be able to export their solutions to a number of public and private actors that face similar climate-related challenges, such as in Southern Europe, California or Australia.

Besides the relevance that these two AI niches hold for Greece, the country is also well-positioned to provide AI compute infrastructure to interested innovators. Greece will host one of the EU-funded AI Factories under the name ‘Pharos’ with the aim to accelerate AI innovation in health, the Greek language and culture and sustainable development. The project will be supported by a budget €30 million, over a period of 36 moths. Greek AI solutions in agriculture or in climate resilience should receive priority access to valuable computational resources through this initiative in order to encourage innovators to develop Greece’s niche.

Greece has the AI talent, and soon will acquire capabilities in AI compute infrastructure. Combined with the necessary reforms in the Digital Single Markets and in Europe’s VC ecosystem, this means that Greek AI innovators have the potential to become developers and even exporters of critical AI solutions. However, the current state of the AI competition forces smaller states to make targeted choices with their limited resources. Casting a wide net will stretch efforts thin and lead to suboptimal results. Instead, Greece should provide incentives for its talent to pursue excellence in one or two AI vertical domains.

Conclusion

AI may drive a significant economic transformation akin to the impact of the steam engine, the electric dynamo and the personal computer. But while past technological revolutions drove prosperity, they were not evenly distributed. The current characteristics of the AI landscape indicate that the European Union risks finding getting the short end of the stick. Without frontrunners that develop frontier AI systems and without businesses and workers that are able and willing to make use of AI, the economic benefits of the technology are likely to flow to the United States and China. Reversing this trend may be one of the most important challenges that Europeans face today. It will require tackling long-standing barriers and thinking creatively about the ways in which AI systems and applications can improve the lives of EU citizens. For an even smaller actor like Greece, this task seems daunting. In order to establish its presence in the AI revolution, the country will need to diagnose its strengths with honesty and hone a critical and unique niche in AI.