Nations are racing to build independent AI capabilities, yet a significant gap persists between ambitious government announcements and the actual technical ability to deliver them. According to research from Stanford's Program on Geopolitics, Technology, and Governance, countries across the EU, Indo-Pacific, and Gulf States are launching sovereign AI initiatives, but many lack clear funding, technical expertise, and a realistic understanding of which AI tasks actually require cutting-edge technology versus cheaper alternatives.

What Exactly Is "AI Sovereignty" and Why Do Countries Want It?

The term "AI sovereignty" sounds straightforward, but it means different things to different nations. Rather than pursuing complete technological independence, which experts consider unfeasible, most countries are actually seeking what researchers call "AI agency" - the ability to make choices aligned with their national interests without depending entirely on U.S. or Chinese technology providers.

The driving force behind this push isn't primarily about safety or shared global risks. Instead, countries are motivated by concerns over economic dependency, market concentration, and geopolitical subordination. Recent events, such as the U.S. government blocking Anthropic's Fable model from certain regions, have reinforced fears that foreign powers could cut off access to critical AI tools at any moment.

Indonesia offers a concrete example of this strategy in action. On August 13, the country officially opened the UGM Indosat NVIDIA AI Technology Center in Yogyakarta, a government-backed hub designed to advance AI research and develop local talent. The facility combines Nvidia's full-stack AI platform with Indosat's sovereign GPU-as-a-service offering, giving Indonesian researchers access to enterprise-grade computing power and pre-trained AI models.

"At Indosat, we believe no Indonesian should be left behind in the AI era. Through UGM Indosat NVAITC, we are bringing the best of global AI technologies and expertise to Indonesia, while expanding access for the ecosystem of researchers, students, startups, and innovators across the country," said Vikram Sinha, president and CEO of Indosat Ooredoo Hutchison.

Vikram Sinha, President and CEO, Indosat Ooredoo Hutchison

Indonesia's Minister of Communication and Digital Affairs, Meutya Hafid, framed the initiative explicitly in sovereignty terms, stating that such projects lay "the foundations of Indonesia's AI sovereignty and ensuring AI becomes a driver of economic growth, national competitiveness, and solutions to Indonesia's most pressing challenges".

How Are Countries Actually Building Sovereign AI?

• Domestic Model Development: Nations are creating their own AI models, often by fine-tuning open-source alternatives to be cheaper, multilingual, and tailored to local markets and use cases that global models ignore.
• National Compute Infrastructure: Countries are building compute clusters and data centers to reduce reliance on foreign cloud providers, with some leveraging low-cost energy or local resources as competitive advantages.
• Data Localization and Access Agreements: Governments are pursuing legal arrangements to keep sensitive data within borders while negotiating assured access to frontier AI models from global providers as a backup plan.
• Talent Pipeline Development: Initiatives like Indonesia's technology center focus on training local researchers and engineers to reduce dependency on foreign expertise.

Indonesia's new AI center is already pursuing practical applications. Initial projects include eNose-TB, an AI-powered screening technology to support the country's fight against tuberculosis, and SmartAgri, which uses AI and precision farming techniques to improve productivity in tropical agriculture.

Why Is There Such a Large Gap Between Plans and Reality?

The Stanford research identifies a critical but often unresolved distinction that undermines many national strategies: determining which AI use cases can rely on "good enough" models using less advanced infrastructure versus those requiring access to frontier capabilities. Many countries are investing billions in sovereign AI projects without clearly answering this fundamental question.

This ambiguity creates real problems. A nation might build expensive compute infrastructure and train engineers to develop advanced models, only to discover that most of its economic and security needs could be met with cheaper, open-source alternatives. Conversely, countries that assume they can always access frontier models from abroad may find themselves vulnerable if geopolitical tensions escalate.

The uncertainty is compounded by a deeper strategic question: will the future of AI be dominated by a handful of frontier labs creating an ever-widening capability gap, or will open-source and fast-follower models remain competitive for most purposes? This fundamental uncertainty is shaping how ambitiously nations are investing in sovereign capabilities.

What Leverage Do Countries Actually Have?

To reduce dependency and create resilience, nations are seeking sources of leverage along the AI value chain. These potential leverage points include control over scarce resources like critical minerals or low-cost energy for data centers, significant market power that makes them indispensable consumers of AI services, and the ability to refine high-value local data for domestic value capture.

However, experts warn that these sources of leverage may be depreciating assets. A nation can typically threaten or use its leverage only once before providers diversify away from it, reducing its long-term negotiating power.

How Are AI-Assisted Cyber Threats Changing the Sovereignty Debate?

While most sovereignty discussions focus on economic and geopolitical concerns, a new threat is emerging that could shift priorities. On August 13, Taiwan disclosed that it had detected AI-assisted cyber-attacks on government agencies originating from overseas in July. The attack used open-source AI agents to build an autonomous hacking tool that behaved like a coordinated cyber team, marking what security researchers described as a first-of-a-kind breach.

The intrusion compromised at least 85 government user accounts and extracted more than 2,500 personnel records before expanding to Taiwan's nuclear safety agency and at least seven energy companies. The attack employed a hybrid approach combining manual operations with AI agent-assisted techniques, according to Taiwan's Ministry of Digital Affairs.

While Taiwan did not formally attribute the attack to China, the use of Simplified Chinese in internal communications linked to the hack suggested a high probability of Chinese involvement. This incident illustrates how quickly AI-assisted hackers can strike and demonstrates a new category of threat that could force countries to reconsider their sovereignty strategies.

"There's still a human in there somewhere. Somebody had to choose who to attack, had to establish an objective and give it a directive. It's not totally 100% autonomous. There was a capable operator in charge that did that," said Cris Thomas, a security researcher and advocate at code security company Semgrep.

Cris Thomas, Security Researcher and Advocate, Semgrep

The threat of AI-assisted hacking campaigns has been growing for years but accelerated dramatically after major AI labs released advanced models capable of rapidly conducting reconnaissance, identifying vulnerabilities, and exploiting them. This development could eventually shift sovereignty discussions from purely economic concerns toward shared security interests, though that transition has not yet occurred in most countries.

What Does This Mean for the Future of Global AI?

The current trajectory suggests a world where nations pursue a mixed strategy: securing assured access to foreign frontier models through diplomatic agreements while simultaneously building domestic capacity as insurance against future cutoffs. However, this approach faces major obstacles, including limited trust in U.S. assurances as a long-term partner and the challenge of coordinating multiple countries around shared standards and infrastructure.

The concept of a "third stack" - an alternative AI ecosystem built on open-source models and diverse hardware by a coalition of middle powers - remains theoretically possible but faces significant collective action problems. Any effort to create an independent alternative could be viewed as a challenge to U.S. national security, potentially incentivizing Washington to pursue bilateral agreements that undermine such efforts.

For now, countries like Indonesia are taking pragmatic steps by partnering with global technology leaders while building local expertise and infrastructure. Whether this hybrid approach successfully delivers on sovereignty promises, or whether the gap between ambition and execution continues to widen, remains one of the defining questions in global AI strategy.