Artificial Intelligence and Geopolitics: An In-Depth Investigation - Report

The era of Artificial Intelligence has arrived and, with it, a radical rearrangement of global powers. As AI promises to unravel the “productivity paradox” and usher in an era of unprecedented progress, a new report reveals an inconvenient truth: China is not only catching up, but redefining technological supremacy. With an impressive research machine, a wave of young talent, and a strategy that makes it the global “primary connector” in AI, Beijing is shaping the future to its advantage. From the “cold war of computation” raging over semiconductors, to military implications and the redefinition of energy supply chains, every move in AI is a geopolitical move. The West is at a crossroads: adapt quickly, invest strategically and forge new avenues for innovation, or risk falling behind in a world where power is increasingly defined by algorithms. Get ready to discover how AI is rewriting the rules of the global game, and why the future may be more Chinese than we imagine.

Credit Extrema Ratio

The People's Republic of China is outpacing the rest of the world in artificial intelligence (AI) research, at a time when this technology has become a "strategic asset" comparable to energy or military capabilities. This is according to a new report titled "DeepSeek and the New Geopolitics of AI: China’s ascent to research pre-eminence in AI", published by Digital Science and edited by its CEO, Dr. Daniel Hook. The analysis reveals that the Asian giant is now the world's dominant power in AI research. Beijing leads the way not only in research volume, but also in citation attention and global influence, an advantage it has rapidly consolidated over the past seven years. In 2024, Beijing's AI publication output equaled that of the United States, the United Kingdom, and the European Union (EU-27) combined, and it now attracts over 40% of global attention in terms of citations. Despite international tensions, the Asian nation has become the leading collaborative partner for the United States, the United Kingdom, and the EU in AI research, although it is less dependent on mutual collaboration than these blocs. Its large and young talent pool fuels sustained research growth, and the launch of models like DeepSeek symbolizes its growing technological independence. While the United States maintains a robust AI startup scene, China is rapidly closing the gap. China also holds a dominant position in AI-related patents, highlighting its ability to transform research into innovation. The United Kingdom, despite its small size, exerts a significant global impact, while the EU-27, while benefiting from strong internal collaboration, faces limitations in external engagement and its translation into patents or commercial activities. AI, with its ability to accelerate research, improve economic productivity, and deliver military advantages, has become an indispensable underlying infrastructure for national strength, impacting security, the economy, and the ability to shape global narratives.

The pervasive impact of AI and its strategic importance

Since the public launch of OpenAI's ChatGPT 3 on November 30, 2022, artificial intelligence has gone from an abstract concept to a tangible and pervasive reality. Generative AI and Large Language Models (LLM) now power widely used chatbots, capable of creating and manipulating text, data, images, and videos with near-magical proficiency, promising to deliver the kind of AI imagined in science fiction. A Digital Science report states that AI is "the most powerful tool humanity has ever created," with its limitations becoming increasingly difficult to imagine given the world's dependence on data and the rise of agentic forms of AI. 

Following the global economic tensions resulting from the 2008 financial crisis and the COVID-19 outbreak, there has been an accelerated need to find ways to increase productivity. Although previous technological revolutions, such as the adoption of computers in the 1980s and the internet revolution in the early 2000s, have not led to widespread increases in productivity, there is strong expectation that AI will eventually fuel the technological revolution that will transform productivity. The Global AI Jobs Barometer 2025 PwC's research supports this forecast, indicating that industries "most exposed" to AI have seen a nearly quadrupling of productivity growth rates, from 7% (2018-2022) to 27% (2018-2024), compared to a decline in less exposed industries..

Goldman Sachs economists have suggested that AI could increase annual productivity growth by between 0.3 and 3.0 percentage points per year over the next decade.

The Digital Science report explicitly references the historical "productivity paradox," presenting AI as intrinsically different due to its "exponential nature" and its ability to "self-replicate" (AI designing and building its own successors). This distinction is crucial because it positions AI as the technology that will finally break the paradox.

Data from 2025 by PwC and Goldman Sachs provides the first empirical evidence to support this assertion, showing significant productivity increases in AI-exposed sectors. This suggests that the impact of AI could indeed be more transformative and sustained than previous technologies, potentially leading to faster and more profound economic restructuring, but also intensifying the associated labor market disruptions.

Deepseek as a case study and symbol of Chinese technological independence

According to Daniel W. Hook, theChina's launch of DeepSeek has challenged the long-held belief that the United States was a decade ahead in AI development. DeepSeek was specifically designed and developed using techniques that circumvented the need for hundreds of millions of dollars of training time that relied on cutting-edge NVidia chips, clearly demonstrating China's impressive AI capabilities.

In addition to matching US restrictions on chips, China has also imposed restrictions on rare earth metals, which could slow down. Producing the same NVidia chips that fueled US dominance. DeepSeek further amplified its impact by releasing its MIT-licensed model as an open-source project, with the stated intent of leveling the competitive playing field.

This affordability and open-source nature are actively attracting global. The Digital Science report explicitly states that the emergence of DeepSeek "is not simply a technological innovation" but "a symbol of a profound change in the global AI landscape." It "exemplifies China's technological independence" and validates the strategy of long-term investment in China's lead in AI.

The open-source release of DeepSeek is presented as an attempt to democratize AI. However, further information reveals a deeper strategic implication: its affordability and open-source availability are leading to its widespread adoption by "multinational banks, universities, and technology companies across Europe, the Middle East, Africa, and Asia.". 

The declaration Microsoft President Brad Smith said, "The No. 1 factor that will define whether the U.S. or China wins this race is whose technology will be more widely adopted in the rest of the world.", provides crucial geopolitical context. This indicates that China is leveraging open-source models not only for technological independence, but as a powerful soft power tool to penetrate global markets, establish technological standards, and in fact build a sphere of influence through broad adoption. This strategy could ultimately lead to a "cold war of calculation" where hardware dominance follows software dominance, fundamentally altering global technological power dynamics.

Dynamic interaction between technological advancement and international power

The Digital Science report emphasizes that AI research, like the space race between the USSR and the United States, is already "fundamentally geopolitical in nature." It states that "whoever controls the best AI will have a competitive advantage in a variety of fields." This advantage extends to an acceleration of research capacity in all scientific fields, not just AI itself. To drive AI, governments must strategically (i) attract or develop and then retain talent; (ii) invest in comprehensive infrastructure, including governance and social structures, to support AI development and implementation; and (iii) promote mechanisms to translate research into broad societal benefits.

The Digital Science report raises a profound question about the evolving nature of global power: "Are these corporations large enough to begin constituting a new level of structure in our world, with their borders drawn in cyberspace rather than physical space?" This directly challenges the traditional model of the nation-state, suggesting that powerful technology companies ("tech nations") like Google, OpenAI, Meta, and Huawei are gaining capabilities and influence traditionally reserved for states. Further analysis notes OpenAI's expansion efforts to maintain dominance and describes Huawei as a "national champion" that is "one of the most powerful in the world." This implies a blurring of the lines between corporate and state power, where tech giants become geopolitical players in fact, potentially creating new forms of governance challenges and power dynamics that transcend national borders and complicate traditional international relations.

The Rise of China's AI: A Comprehensive Power Ploy

• Chinese research at the forefront: volume, talent and influence

According to Digital Science, theGlobal AI research has grown impressively since 2000, with an average compound annual growth rate of 8%, reaching 60,000 publications in 2024. Beijing's contribution to this growth has been exceptional, with its production volume in 2024 equaling the combined output of the EU-27, the United Kingdom, and the United States. This signifies a "massive and impressive" expansion of China's AI research capacity.

China's AI research ambition is further demonstrated from its vast and uniquely structured talent base. With approximately 30,000 AI researchers of all ages, China's AI community significantly surpasses that of the EU-27 (nearly 20,000) and the United States (about 10,000). A distinctive feature is China's large and young population of AI-trained researchers, where the combined populations of doctoral students and postdoctoral fellows are nearly double the total AI population of the United States.

In terms of influence, the Asian nation is constantly predicted to become the most influential country in global AI research in 2018, measured by eigenvector centrality, surpassing both the EU-27 and the United States. This dominance extends to research visibility, with Chinese output receiving over 40% of the global attention available for citations in 2024.

According to Daniel W. Hook, theChina's large pool of young AI-trained talent, particularly its doctoral and postdoctoral populations, represents a profound "demographic dividend" for its AI ambitions. Unlike Western nations, which exhibit a more balanced age distribution among researchers or rely on talent imports, China's "almost nonexistent" older cohorts and limited large-scale talent imports among these groups. They indicate a self-sustaining and domestically cultivated talent base. This "young, vibrant, and highly educated AI-savvy workforce" is explicitly connected fueling "a wave of innovation from China along the lines of DeepSeek." This unique demographic structure provides a long-term and sustainable advantage in AI development, as it ensures a continuous flow of innovators and leaders who deeply understand the technology, driving internal activity and innovation at an unparalleled scale.

Year	China	UE-27	United Kingdom	United States	Global Total
2000	<1000	~1000	<500	~2000	~10000
2005	~2000	~2000	~1000	~3000	~15000
2010	~5000	~4000	~2000	~6000	~25000
2015	~10000	~6000	~3000	~8000	~35000
2020	~18000	~10000	~4000	~10000	~50000
2024	~24000	~12000	~5000	~10000	~60000

Table 1: Global AI research output by key region (2000–2024). Approximate data based on Digital Science report, July 2025.

Age range	China (2020)	UE-27 (2020)	United Kingdom (2020)	United States (2020)
Student (0-4 years)	~12000	~5000	~1000	~3000
Postdoc (5-9 anni)	~8000	~4000	~800	~2500
Early career (10-14 years)	~4000	~3000	~600	~2000
Established researcher (15-19 years old)	~2000	~2000	~400	~1500
Career Builder (20-24 years old)	~1000	~1500	~300	~1000
Peak production (25-29 years)	~500	~1000	~200	~800
Advanced Researcher (30-34 years)	~200	~500	~100	~500
Senior Researcher (>35 years)	<100	~300	~50	~300
Approximate total	~27800	~17800	~3450	~11600

Table 2: Demographics of AI researchers by region (2020 snapshot). Approximate data based on Digital Science report, July 2025.

• The role of "primary connector" in international collaboration on AI

Despite recent trends towards deglobalization and fragmentation, China has consolidated its position as a "primary connector for AI research" globally. The UK's collaboration patterns illustrate this shift dramatically: historically, the US was a high-volume collaborator, but China's collaboration with the UK increased from just 2% of UK output in 2000 to over 25% by 2024, making China the UK's largest AI collaborator. This means that the UK is now "far more dependent on Chinese than US collaboration in AI".

From the EU-27 perspective, while maintaining strong internal cooperation, China has recently surpassed both the UK and the US as a major AI collaborator, despite the EU-27's weaker overall external dependencies. From the US perspective, collaboration with China is also growing consistently through 2019, with China remaining the United States' most regular research collaborator in AI, accounting for 14% of U.S. production. This is happened despite subsequent US federal actions such as the China Initiative (2018) and the Securing American Science and Technology Act (2019), China's dependence on collaboration with the EU-27, the UK, and the US is remarkably low, at 4% in 2024, compared to significantly higher rates for these regions' dependence on China.

Year	China to the USA	China to UK	China versus EU-27	USA towards China	UK to China	EU-27 versus China
2000	Net Inflow	Net Inflow	Net Inflow	Net Outflow	Net Outflow	Net Outflow
2005	Net Inflow	Net Inflow	Net Inflow	Net Outflow	Net Outflow	Net Outflow
2010	Net Inflow	Net Inflow	Net Inflow	Net Outflow	Net Outflow	Net Outflow
2015	Net Inflow	Net Inflow	Net Inflow	Net Outflow	Net Outflow	Net Outflow
2020	Net Inflow	Net Inflow	Net Inflow	Net Outflow	Net Outflow	Net Outflow
2024	Net Inflow	Net Inflow	Net Inflow	Net Outflow	Net Outflow	Net Outflow

Table 3: Net AI talent flows (selected regions, 2024). This table represents a qualitative trend based on the Digital Science report, July 2025.

The persistent and growing cooperative ties between China and Western nations, particularly the emergence of China as a "primary connector" are surprising given broader geopolitical tensions and trends toward deglobalization. This suggests that scientific collaboration, especially in a critical and rapidly evolving field like AI, can act as a powerful form of soft power that transcends political divides. As a central node in international research networks, China not only accelerates its own research and talent development, but also subtly influences global research agendas, fosters dependence on its own scientific ecosystem, and gains insight into the directions of international research. This dynamic highlights a tension between the scientific community's desire for open collaboration ("researchers may not want to see international borders") and the international community's desire for open collaboration ("researchers may not want to see international borders" states Digital Science) and national strategic interests, creating a complex web of interdependence that Beijing can potentially exploit for long-term strategic advantage.

• Innovation delivery: patents, corporate involvement, and broad geographic expertise

China's vast talent base translates directly into significant innovation, particularly evident in its patenting trends. In recent years, the growth of AI patents in China is expected to accelerate dramatically, with China already producing nearly ten times more patents than its closest competitor. This remarkable effect is not due to just a few centers of excellence, but to a widespread capacity.

China also leads the way in university-business collaborations, as evidenced by the high number of publications involving corporate entities. Although the United States has historically had a strong technology sector, China is rapidly catching up in terms of actively publishing corporate entities, suggesting it is poised to surpass the United States in AI startups engaging in research.

Year	China	United States	Japan	Germany	United Kingdom	South Korea	Canada
2000	~10	~100	~50	~20	~10	~10	~5
2005	~30	~200	~80	~30	~20	~20	~10
2010	~100	~300	~100	~50	~30	~30	~15
2015	~1000	~1000	~200	~100	~50	~50	~20
2020	~10000	~3000	~500	~200	~100	~100	~30
2024	~30000	~5000	~1000	~300	~150	~150	~50

Table 4: AI patents filed by key countries (2000–2024). Approximate data based on Digital Science report, July 2025.

The geographic distribution of AI expertise in China is particularly striking. Unlike other key geographies where AI research tends to be more concentrated, China has amassed a significant volume and density of AI capabilities across the country. Focus points include Beijing, Shanghai, Nanjing, and Guangzhou, with 156 organizations producing more than 50 AI-related papers in 2024, a stark contrast to the 54 organizations in the EU-27 that met the same criteria.

The widespread geographical distribution of AI expertise and high-production research organizations across China is a key factor of differentiation. The report explicitly contrasts it with the more concentrated hubs in the EU and the United States. This "relatively evenly distributed" research suggests a deliberate national strategy to foster innovation in numerous centers, rather than concentrating resources in a few "golden triangles." This decentralized model improves national resilience by distributing capabilities, accelerates diverse applications by fostering local innovation ecosystems, and creates a broader base for talent deployment and corporate spin-offs. This approach aligns with the concept of "choosing competition rather than picking winners", making the overall innovation ecosystem more robust and less susceptible to single points of failure, which is a significant strategic advantage in a rapidly evolving field like AI.

Strategic initiatives and military AI integration (PLA's generative AI, chip independence)

China's People's Liberation Army (PLA) has embarked on a aggressive action to integrate generative AI to revolutionize its capabilitiesintelligenceThis involves the use of advanced Large Language Models (LLM) for critical tasks such as data processing, analysis of the intelligence and decision support the rapid adoption of DeepSeek models in early 2025 highlights the PLA's aggressive push to employ AI across its entire research workflows intelligence. Chinese defense contractors have even affirmed to provide the PLA with models of open-source intelligence (OSINT) custom based on DeepSeekThis demonstrates a direct link between civilian AI innovation and military applications.

In response to US export controls on high-performance semiconductors, Chinese tech giants like Huawei and Semiconductor Manufacturing International Corporation (SMIC) are racing to build home alternatives Huawei's Ascend series, especially the Ascend 910C, is emerging as a significant competitor to NVIDIA's H100, offering inference capability comparable at a lower cost and is scheduled for mass production in 2025 Huawei's global distribution network, built on its presence in telecommunications, is now selling actively AI infrastructure, including locally produced GPUsChina's 14th Five-Year Plan for Digital Economy Development explicitly mandates the overseas expansion of digital infrastructure, including data centers and AI computing resources, to support Chinese digital enterprises in "go global." This is a deliberate strategy to export its computing infrastructure and standards.

The Cyberspace Administration of China (CAC) has also taken measures to manage AI-generated content nationwide, releasing measures for the identification of AI-generated synthetic content, effective from 1 September 2025. That indicates a focus on information control and combating the potential misuse of AI-generated disinformation.

The direct integration of civilian-developed generative AI models, such as DeepSeek, in operations intelligence PLA military, vividly illustrates the profound nature dual-use of AI. It is not simply about developing advanced military hardware, but about exploiting commercial AI breakthroughs for a strategic advantage in areas such as the analysis of intelligence and the information war. The simultaneous push toward chip independence and the global export of Chinese computing infrastructure reveal a comprehensive strategy to control the entire AI value chain, from fundamental hardware to advanced applications. This implies that the development of AI, including in the civilian sector, has immediate and significant implications for national security, necessitating a reassessment of how open-source models and civilian AI research are managed and regulated in a geopolitical context, as the "cold war of computation"encompasses the entire AI ecosystem.

Tabella: AI growth metrics in China (2000-2024/2025)

Metrics	2000	2024/2025	Details and Context
AI research output (publications)	<10,000 (global) (from Digital Science report, July 2025)	60,000 (global) (from Digital Science report, July 2025); China = EU-27 + UK + US combined (from Digital Science report, July 2025)	China has demonstrated a massive and impressive expansion of its AI research capacity, equaling the combined output of major Western powers.
AI Talent Pool (Active Researchers)	Not specified	~30,000 (China) vs. ~20,000 (EU-27) vs. ~10,000 (US) (from Digital Science report, July 2025)	China has a young and large AI population; Chinese doctoral and postdoc students are nearly double the total AI population of the United States (from Digital Science report, July 2025).
Global attention share (citations)	Single digits (China) (from Digital Science report, July 2025)	>40% (China) (from Digital Science report, July 2025)	Chinese manufacturing has received over 40% of available global citations since 2024, demonstrating a growing influence on research visibility.
AI patents filed	Low (China) (from Digital Science report, July 2025)	Nearly 10 times more than the nearest competitor (China) (from Digital Science report, July 2025)	AI patent growth in China has accelerated significantly, indicating a significant translation of research into innovation.
Business entities that actively publish	Not specified	China is rapidly approaching the level of the United States (from Digital Science report, July 2025)	China is leading the way in university-business collaborations and is closing the gap with the United States in AI startups engaging in research.
Main collaborators (collaboration share)	China: ~2% with UK (from Digital Science report, July 2025)	UK: China > 25% (main contributor) (from Digital Science report, July 2025); EU-27: China has overtaken UK and US (from Digital Science report, July 2025); US: China ~14% (most regular contributor until 2019) (from Digital Science report, July 2025)	China has become the “primary connector” for AI research, with low mutual dependence (4% for China) compared to its Western partners (from Digital Science report, July 2025).

Western AI Strategies: Adaptation and Competition

• The United States: Variable Influence, Enduring University-Industry Links, and Talent Migration Patterns

According to Digital Science, the United States continues to demonstrate leadership in translating AI research into innovation, primarily through its strong university-industry ties. This is reflected in the number of articles published by corporate entities, where the United States remains a significant player, although China is rapidly catching up. However, the report notes that much of the innovation in the United States, particularly from private research organizations like OpenAI, may not be fully captured by traditional publication metrics.

Despite its historical leadership, the United States' overall influence in AI research is in decline, with China now holding a dominant position in terms of research output. While the United States remains an attractive destination for AI researchers globally, it is experiencing significant net talent flows and "overwhelming" towards China. This loss of talent is a significant concern for its long-term innovation capacity.

In terms of collaboration, the US's AI link with the UK is only doubled between 2000 and 2024 (5% to 10% of UK-based articles), despite the major UK-based AI companies being US-owned. From the US perspective, its AI collaboration with other nations appears to have peaked around 2012 and has declined since then, falling to nearly 10%. However, China remained the largest research contributor to the United States in AI through 2019, representing 14% of U.S. production.

The United States federal government has launched several initiatives to strengthen its AI capabilities, including the National AI Research Resource pilot to democratize access to AI, the AI ​​Talent Surge to accelerate the hiring of AI professionals, and the EducateAI initiative to fund educational opportunities in AI.Legislative efforts reflect growing geopolitical tensions: the "Decoupling America's Artificial Intelligence Capabilities from China Act of 2025" (S. 321) was introduced to prohibit U.S. persons from promoting AI capabilities in China and prohibit the import/export of AI/generative AI technology or intellectual property to/from China. On the contrary, the Trump administration has removed some restrictions on the use and development of AI by early 2025 and has repealed the Biden-era "AI Diffusion Rule" to "unleash American innovation and ensure American dominance in AI".

The Digital Science report identifies the United States as a global destination for AI talent, which is a clear strength. However, the significant and overwhelming net loss of AI talent to China is a vulnerable strategic criticism. This outflow of human capital directly addresses one of the three key imperatives for AI leadership: the ability to "attract or develop and then retain talent." This "brain drain" could erode the United States' long-term innovation capacity, particularly when compared to China's massive and growing domestic talent base. This suggests that, despite strong university-industry ties and domestic initiatives, the United States faces a fundamental challenge in retaining its best AI minds, which could have cumulative negative effects on its competitive advantage.

• National AI initiatives and the evolution of military AI doctrines (cybersecurity, autonomous systems)

The United States Department of Defense (DoD) is actively integrating AI to accelerate its decision-making processes in combat, incorporating AI into targeting systems, predictive analytics, and unmanned combat operationsA key strategic imperative is to align research and development (R&D) priorities to meet the growing demand for autonomous warfare capabilities.

The National Defense Authorization Act (NDAA) for Fiscal Year 2025 provides priority research into emerging AI and quantum computing technologies to improve defense operations. Establishes pilot programs to evaluate AI development for security-related biotechnology applications and workflow optimization at DoD facilities. The NDAA authorizes even the DoD to use AI for internal business operations, such as auditing financial statements. The duties of the Chief Digital and AI Officer Governing Council have been expanded to identify and assess AI models that could pose national security risks if accessed by adversaries, and to develop strategies to prevent unauthorized access and use of powerful AI models by hostile nations.

The cybersecurity landscape is evolving rapidly, with approximately 40% of all cyberattacks now driven by AI. Attackers using AI are increasingly being used to create adaptive malware and automated phishing attempts. A new and emerging threats it's "AI model manipulation," where attackers shift their focus from stealing data to "poisoning the AI ​​models themselves."The experts they foresee that by 2025, this could escalate into "full-scale machine warfare," requiring highly sophisticated security operations centers capable of making complex tactical decisions at "machine speed."

Recognizing the significant energy needs of AI, the Bipartisan Policy Center (BPC) Task Force on AI and Energy was launched in April 2025. This task force focuses on AI electricity demand, grid reliability, and clean energy solutions for data centers, supporting DOE investments in energy-efficient AI systems and grid resilience..

The aggressive integration of AI in US military operations and the explicit attention to identify and mitigate risks arising from "AI-using adversaries", clearly signal an intensification of the arms race in AI. The emergence of "AI model manipulation" as a new threat vector and the prediction of a "full-scale machine war" in cybersecurity by 2025, indicate a significant escalation in the nature of conflict. This implies that future conflicts will increasingly be fought not only with physical weapons, but with autonomous AI systems engaging in digital combat, potentially at speeds beyond human comprehension. This necessitates a fundamental reevaluation of traditional defense strategies, emphasizing robust AI cybersecurity, proactive threat intelligence, and the development of advanced AI-based defense capabilities to counter adversarial AI.

• The United Kingdom: collaborative strengths, global impact, and deepening ties with China

Second Digital Science, il United Kingdom It maintains a highly collaborative research economy, strategically leveraging international partnerships to preserve its soft power through research. Despite its relatively smaller research economy (around 3,000 active AI researchers), the UK consistently demonstrates global impact, outperforming the average in terms of "attention per output" in AI publications, suggesting high-quality research.

A significant change in the UK's collaborative landscape is the deepening of its ties with China. China has become the UK's largest AI collaborator, surpassing both the US and the EU. Collaboration with China is increased from just 2% of UK output in 2000 to over 25% by 2024. This growing dependence means that the UK is now “far more dependent on Chinese than US collaboration in AI.”

In terms of talent flow, while the UK has historically attracted researchers from the EU-27, more recently it has become a net donor of AI talent to China.

The UK's position as a highly collaborative research economy and its increasingly deep and dependent ties with China in AI, while maintaining links with the US and EU, position it as a potential "bridge" between different geopolitical blocs. However, this role entails a significant geopolitical dilemma. Over-reliance on a rising strategic competitor like China, coupled with a net outflow of talent to China, could compromise strategic autonomy of the United Kingdom and expose it to risks associated with the previously identified "asymmetric interdependence." This raises critical questions about how the UK balances the economic and scientific benefits of collaboration with national security concerns and alignment with its traditional Western allies.

• National Action Plans on AI and Defense Modernization Efforts (Digital Targeting Network, Offensive Cyber ​​Defense)

The UK AI Opportunities Action Plan 2025, launched on January 13, 2025, outlines a comprehensive strategy to position the UK as a global leader in AI. Key themes include promoting economic growth, ensuring ethical development of AI, investing in skills development, improving digital infrastructure (including ensuring computing power and developing "AI Growth Zones") and promotion of public trust and cooperation.

The Strategic Defence Review 2025 (SDR)fundamentally redefines the role of AI in the UK's defence, viewing it not as a peripheral tool but as a fundamental component of modern warfare. SDR aims to transform the UK Armed Forces into a technology-enabled Integrated Force, improving speed of decision-making and supporting autonomous systems. A critical objective of the SDR is the implementation of a "digital targeting network" by 2027, with a minimum functioning product by 2026. This network will exploit AI for real-time data analysis, operational strategy optimization, and decision-making acceleration by connecting sensors, decision makers, and effectors across different domains.

The SDR announces the creation of a new Cyber ​​and Electromagnetic (CyberEM) Command will also be announced. This command will be responsible for overseeing the integration of AI into cyber operations, electromagnetic warfare, and information operations. He will drive cyber defence operations and will coordinate offensive cyber capabilities with the National Cyber ​​Force, supported by an investment of over £1 billion. L'I commit of the United Kingdom in developing a "digital targeting network" and establishing a new "CyberEM Command"demonstrates a proactive and integrated approach to national defense in the AI ​​era. The explicit mention of the coordination of "offensive cyber capabilities" signifies a critical evolution in military doctrine, recognizing that modern warfare increasingly involves digital aggression and counter-aggression. This blurring of the lines between offensive and defensive in the cyber domain, driven by AI, implies a shift toward continuous engagement and the need for constant innovation and adaptation to maintain a strategic advantage. It underscores that national security now depends as much on digital supremacy as on traditional military power.

Tabella: comparative strengths and challenges of AI

Metrics	United States	United Kingdom	Details and Context
Overall influence of AI research (eigenvector centrality)	In decline since 2018 (from Digital Science report, July 2025)	More or less preserved (from the Digital Science report, July 2025)	China has surpassed both of them in overall influence.
Links between academia and industry	Continue to lead the way in AI-innovation translation (from the Digital Science report, July 2025)	Strong, but not strongly correlated with partners (from Digital Science report, July 2025)	Much of U.S. innovation is in private contexts not captured by metrics.
Talent flows	Crushing net loss to China (from Digital Science report, July 2025)	Net donor to China (from Digital Science report, July 2025)	Both countries are losing AI talent to China.
Main collaborators	China (until 2019, 14% of US output) (from Digital Science report, July 2025)	China (over 25% of UK output, main contributor from 2024) (from Digital Science report, July 2025)	The UK is much more dependent on Chinese AI collaboration than the US.
Focus of the National AI Strategy	Democratizing access, hiring professionals, financing education (Chambers and Partners, 22 May 2025, available athttps://practiceguides.chambers.com/practice-guides/artificial-intelligence-2025/usa)	Economic growth, ethical AI, skills development, digital infrastructure, public trust (SCC, January 16, 2025, available athttps://www.scc.com/insights/it-solutions/a-guide-to-the-2025-uk-ai-opportunities/; Crowe UK, 2025, available onhttps://www.crowe.com/uk/insights/uk-ai-opportunities-action-plan)	The US focuses on access and talent; the UK on a broader strategy.
Military AI Integration	Decision cycle acceleration, autonomous systems, pilot programs (biotech, workflow), adversarial risk mitigation (Frost & Sullivan, 2025, available athttps://www.frost.com/growth-opportunity-news/aerospace-defense/defense-industry/redefining-u-s-defense-strategic-imperatives-for-2025-cim-ps/; K&L Gates, January 2, 2025, available at(https://www.klgates.com/Key-Provisions-on-Artificial-Intelligence-in-Fiscal-Year-2025-NDAA-1-2-2025))	Digital Targeting Network (by 2027), CyberEM Command (offensive/defensive), investment > £1bn (Burges Salmon, 11 June 2025, available athttps://www.burges-salmon.com/articles/102kdtq/ai-and-defence-insights-from-the-strategic-defence-review-2025/; GOV.UK, 29 May 2025, available athttps://www.gov.uk/government/news/uk-to-deliver-pioneering-battlefield-system-and-bolster-cyber-warfare-capabilities-under-strategic-defence-review; The Defense Post, May 30, 2025, available athttps://www.thedefensepost.com/2025/05/30/uk-cyber-targeting-network/)	The United States is focused on optimizing and mitigating risks; the United Kingdom is focusing on operational transformation and offensive cyber capabilities.

The UE: Balancing Ambition and Reality

• Internal collaboration vs. external engagement and gaps in innovation translation

Second Digital Science, l'EU-27 demonstrates significant strength in internal collaboration on AI across its research blocs, fostering an "intrinsic richness" within its scientific community. However, this internal focus is accompanied by weaker external dependencies. The EU-27 collaborates on a maximum of 8% of its AI research outside its borders. This limited external engagement is identified as a potential obstacle to the “EU-27 innovation dynamism in AI”.

Despite its internal strengths, the EU-27 "underperforms in the visibility of AI research and its translation into patents or business." While it gets a similar share of global attention as the United States (about 10%), when normalized for output volume, the EU-27 consistently receives less attention compared to its production compared to China, the United States and the United Kingdom.

The combination of strong internal collaboration within the EU-27 and its acknowledged "underperformance in AI research visibility and translation into patents or business", indicates a significant "innovation chasm." This suggests that, while the EU excels in fundamental research and internal knowledge exchange, it struggles to convert this academic excellence into tangible economic value, commercial products, or influence on the global market. The report explicitly warns that this could lead to a "vicious rather than virtuous cycle" in which limited translation opportunities discourage talent acquisition. This implies a systemic challenge in bridging the gap between scientific discovery and industrial application, potentially due to regulatory complexities, a fragmented single market, or cultural differences in risk-taking versus more agile innovation ecosystems.

Year	China	United States	UE-27	United Kingdom	Others
2000	~5%	~40%	~15%	~5%	~35%
2005	~10%	~35%	~15%	~5%	~35%
2010	~20%	~30%	~15%	~5%	~30%
2015	~30%	~20%	~10%	~3%	~37%
2020	~35%	~15%	~10%	~2%	~38%
2024	~40%	~10%	~10%	~2%	~38%

Table 5: Global attention share of AI citations by region (2000–2024). Approximate data based on Digital Science report, July 2025.

Year	China	United States	UE-27	United Kingdom	Others
2000	~0.5	~1.5	~0.8	~1.8	~0.9
2005	~0.8	~1.3	~0.8	~1.6	~0.9
2010	~0.9	~1.2	~0.9	~1.5	~0.9
2015	~1.0	~1.1	~0.9	~1.4	~0.9
2020	~1.1	~1.0	~0.9	~1.3	~0.9
2024	~1.2	~1.0	~0.9	~1.2	~0.9

Table 6: Normalized voice share in AI research (2000–2024). Approximate data based on Digital Science report, July 2025.

• EU AI Act: Implementation, Prohibited Practices, and Regulations on General-Purpose AI (GPAI)

The EU AI Act', recognized as the world's first comprehensive legal framework for artificial intelligence, officially entered into force on 1 August 2024 the first substantive obligations, in particular the provisions on "prohibited AI practices" (e.g. manipulative AI techniques, exploitative systems targeting vulnerable groups, real-time remote biometric surveillance in public spaces with limited exceptions), have become applicable in February 2025. Organizations are now also responsible to ensure that personnel interacting with AI systems have adequate AI literacy.August 2, 2025 marks a key milestone, as introduce extended obligations for general purpose AI (GPAI) models and activates new frameworks governance, including the European AI Office and the European Artificial Intelligence CommitteeGPAI model providers, particularly those offering Large Language Models (LLM), will face new horizontal obligations related to transparency, documentation, and copyright compliance. For GPAI models deemed to pose systemic risk, they will apply additional requirements such as risk mitigation, incident reporting and cybersecurity safeguardsRecital 110 of the AI ​​Act explicitly lists the "systemic risks" that GPAI providers they must identify and mitigate, including CBRN misuse, offensive cyberattack capabilities, self-replicating patterns, and large-scale disinformationThe broader picture, including obligations for most high-risk AI systems, will enter in full force on August 2, 2026, with the final deadline for the compliance set for August 2, 2027.

However, as of May 2025, they are emerged reports suggesting that the European Commission may postpone the application and implementation of some provisions (e.g. high-risk AI systems, due on 2 August 2026) due to the lack of technical standards, requests for exemptions for SMEs and derogations for low-complexity AI systems. This is part of a broader effort to "simplify" EU legislation. The EU's leadership in AI regulation, while establishing a comprehensive legal framework, could create a more cautious environment that inadvertently hinders innovation dynamism and translation, compared to more agile or less regulated ecosystems. Ongoing discussions on postponing or simplifying the AI ​​Act indicate a domestic recognition of the potential for overregulation to impact competitiveness, creating a tension between the EU's regulatory ambitions and its innovation goals.

• Strategic programs: AI Continent Action Plan, GenAI4EU, and AI Factories

The overall EU strategy on AI aims to establish the Union as a world center of excellence for AI, ensuring that AI development is human-centric and trustworthy.

A key initiative is theAI Continent Action Plan, launched in April 2025, which aims to position Europe as a global leader in AI. This piano focuses on developing trustworthy AI technologies to improve Europe's competitiveness while safeguarding and promoting democratic values. It seeks to bring the benefits of AI to various sectors, including healthcare, education, industry, and environmental sustainability.The plan include actions to: build large-scale AI data and computing infrastructure; increase access to high-quality data; promote AI adoption in strategic sectors; strengthen AI skills and talent; and facilitate the implementation of the AI ​​Act.The AI ​​Continent Action Plan also aims to accelerate the adoption of AI in strategic sectors such as healthcare, automotive and advanced manufacturing, and supports companies and public administrations in the development and implementation of promising AI solutions.

GenAI4EU is a landmark initiative within the "Communication to stimulate startups and innovation in trustworthy artificial intelligence". Its scope is to stimulate the adoption of generative AI in key strategic industrial ecosystems in the Union. It will also encourage the development of large, open innovation ecosystems that foster collaboration between AI startups and AI users in industry and the public sector..

The establishment ofAI Factories and GigafactoriesIt is a key component of the AI ​​Continent Action Plan, aimed at building large-scale AI computing and data infrastructure.The EU provides to mobilise €20 billion for AI infrastructure, aiming for up to five AI Gigafactories across the EU – large-scale facilities to develop and train complex AI models on an unprecedented scale, integrating massive computing power of more than 100,000 advanced AI processors.The EU aims to mobilise investments significant, with the Horizon Europe and Digital Europe programmes investing €1 billion per year in AI, and further private and Member State investments targeting an annual volume of €20 billion over the course of the digital decade.

The AI ​​Act full of the EUand his strategic programs such as the AI ​​Continent Action Plan,demonstrate a deliberate strategy to drive the governance of AI and ethical development. This "regulatory first" approach, which prioritizes safety, fundamental rights, and reliability, differs from the innovation-driven strategies of the United States and China. The EU's efforts to establish global standards, as evidenced by the ISO/IEC 42006:2025 standard, suggest that its regulatory framework has extraterritorial ambitions. This means that companies operating globally, even if not based in the EU, may be required to comply with EU standards to access the vast European market. This approach, while potentially creating "innovation friction" internally, allows the EU to exert significant "regulatory power" over the global AI landscape, shaping the way AI is developed and implemented worldwide by influencing international standards and practices.

• Military initiatives on AI and infrastructure (European Defence Fund, PESCO, transport)

The EU Strategic Compass for Security and Defence underlines the growing importance of innovation in defence, recognising its strategic value and emphasising the need to strengthen the EU's emerging military technologies, including AIOver the last decade, the EU and its Member States have increased investments in AI-based military technologies.

Multiple projects of theEuropean Defence Fund (EDF)and of the Permanent Structured Cooperation (PESCO) are dedicated to the integration of AI into future military capabilities.

Digital technologies, including cybersecurity and the use of AI to enhance military capabilities, feature heavily in this plan, reflecting the importance of protecting critical infrastructure and military networks from digital threatsThe EDF also emphasizes the integration of AI into defense systems to improve efficiency and responsiveness to emerging threats, and aims to reduce fragmentation investments and promote interoperability between Member States.

In the transport sector, although the EU AI Act mentions "safer and cleaner transport" as a benefit of AI, the EU's investment of €2.8 billion in 94 transport projects in 2025 is focused extensively on sustainable and connected mobility through the Trans-European Transport Network (TEN-T). These projects include modernization of railways (e.g., Rail Baltica), improving digital traffic management in countries such as France and Spain, and developing cooperative intelligent transport systems (C-ITS) for road safety. Although these systems promote “smarter” and “connected” mobility, the sources do not explicitly specify the role of AI in these civil infrastructure projects beyond the general implications of digitalization or traffic management.

Conclusions: Navigating the AI-Driven Geopolitical Landscape

Digital Science's report "DeepSeek and the New Geopolitics of AI" paints an unequivocal picture: artificial intelligence is no longer a neutral scientific discipline, but a key strategic asset reshaping the global balance of power. China has established itself as the undisputed leader in AI research, surpassing Western powers in terms of production volume, citation influence, and, crucially, its ability to translate research into practical innovation, as demonstrated by its patent prominence and the emergence of models like DeepSeek. Its large and young talent pool, combined with a distributed research strategy and an asymmetric international collaboration model, gives it a long-term competitive advantage.

The United States, while maintaining a strong startup ecosystem and university-industry ties, faces the challenge of declining research influence and a significant brain drain of AI talent to China. Its strategy is increasingly focused on technological decoupling and strengthening AI-based military capabilities, but persistent research interdependence and the global spread of Chinese models complicate this path. The United Kingdom, with its high-quality research and collaborative approach, is uniquely positioned as a potential bridge between East and West, but must navigate its growing dependence on Chinese collaboration and mitigate the outflow of talent. The European Union, while excelling in internal collaboration and attempting to establish a comprehensive and ethical regulatory framework with the AI ​​Act, risks lagging behind in innovation translation and external engagement, crucial factors for competitiveness.

The geopolitical dynamics of AI extend beyond research, touching on the vulnerabilities of semiconductor supply chains, AI's enormous and growing energy needs, and its impact on cybersecurity and information warfare. The global race for national AI champions and the development of international standards reflect a broader competition for technological dominance and regulatory influence. While AI offers unprecedented opportunities to improve productivity and address societal challenges, it also risks exacerbating existing inequalities, with developing nations struggling to access the necessary talent and resources.

Ultimately, AI is poised to shape the global order of the 21st century. China has demonstrated a long-term, multifaceted strategy that is bearing fruit, while Western nations must adapt rapidly, not only with reactive policies, but with strategic and sustained investments in talent, infrastructure, and innovation to remain competitive. The real challenge will not just be who develops the most powerful AI, but who can most effectively integrate it into their economies and security, while managing its profound global social and ethical implications.