Research
Export controls have materially reshaped the AI landscape — cutting US semiconductor sales to China by $22.4B while accelerating Chinese domestic production by 310% — but whether sovereign firewalls ultimately accelerate or retard innovation remains genuinely contested, with evidence suggesting China's firewalled model iteration cycle (14 days) outpaces Germany's open-integrated one (22 days), complicating the standard assumption that openness drives speed. The deeper tension is not security versus openness but who the "commons" actually serves: 87% of top AI conference papers originate in the Global North, under 0.3% of Hugging Face models are maintained by African institutions, and the Foreign Direct Product Rule extends US semiconductor jurisdiction extraterritorially, making "open research" function, for many Global South actors, as managed dependency rather than shared infrastructure. Innovative sovereignty models from Korea (mutual model reciprocity), India (provenance-tagged federated commons), and ASEAN (tiered sovereignty architecture) suggest the binary of firewall versus openness is already obsolete — but longitudinal data on whether fragmentation degrades AI safety coordination, and honest accounting of compute costs for researchers outside OECD institutions, remain absent from the literature.
Sovereign AI is a growing trend that sees states pursue their own interests by investing in domestic AI capabilities and expertise. Image ...
... EU, and introduce a single EU-wide framework to assess cloud and AI sovereignty; strengthening digital autonomy through open source – the ...
AI firewalls are security controls designed to protect generative and predictive AI applications, models, agents, prompts, data flows, and connected tools.
For example, the October 2022 controls introduced measures banning the sale of high-end AI chips to any entity operating in China.
Sovereign AI blocs are emerging as governments worldwide spend billions of dollars subsidizing domestic computing infrastructure and AI models.
This report analyzes the potential impact of high-end consumer GPUs on the efficacy of US export controls on AI chips.
Building on commons scholarship, OpenAI is benchmarked against the class of entities that AI experts have suggested as a more trustworthy alternative, i.e.,
This study identifies key policy priorities, emerging technological solutions, and best practices to ensure that open-source AI remains a force ...
UK's rebranding of its AI Safety Institute to the "AI Security Institute" in February 2025.
The EU AI Act was published in the Official Journal (OJ) of the European Union on 12 July 2024. This is the final version of the text to be enforced.
The AI Research Commons Midwest Fellows Program is a highly selective program connecting high-potential, inception-stage AI startups from leading Midwest ...
The final section considers potential avenues for researching AI nationalism, elaborating upon the potential causal mechanisms for its emergence ...
The BAIR Open Research Commons (“BAIR Commons”) is an industrial affiliate program designed to accelerate cutting-edge AI research. AI research is advancing ...
There is a “trilemma” among global openness, digital sovereignty and the risk of data colonialism. For researchers in the Global South ...
The global race for AI supremacy is already contributing to the formation of AI blocs and the fragmentation of shared technological standards.
While the development of 'sovereign AI' appears contradictory, it might have advantages for Russian FIMI campaigns. Namely, a more ideologically ...
China's accelerating wave of AI regulation makes one thing clear: the era of lightly-governed open-source AI in the country is ending.
Russia is pursuing a sovereignty-first, state-directed AI strategy to strengthen national security and economic resilience through 2030, ...
Eligible users can access AI compute at up to 40% reduced cost under the IndiaAI Mission. Mission is developing a scalable AI computing ecosystem to support ...
New Legal Directions for a Global AI Commons. Spring Speaker Series. Ethics and Governance of AI Technology & the Law.
The global concentration of compute infrastructure makes open-weight models one of the most viable pathways to sovereign AI capacity in the Global South; restricting such access deepens asymmetries while driving proliferation into unsupervised settings. ... Sovereign AI development requires access...
On the other, osAI risks catastrophic misuse, threatens national security, and will inevitably be co-opted by dominant corpo- rate interests, reinforcing the very power structures they claim to challenge.20 ... Adherence to open science principles—transparent peer review, public dissemination of r...
|Rank|Country|Year|Strategic Framework Title|Governance Domain|Performance Score|Verified Asset Link| ... |Country|Primary Supercomputing Node|Infrastructural Configuration / Hardware|Status / National Mission Alignment| ... |Brazil|Santos Dumont (LNCC)|Multi-petaflop CPU/GPU hybrid environment, Pet...
Based on previous literature, we propose to break these questions down to three levels: (1) how much AI compute a country has on its territory, (2) what is the nationality of the companies who own the AI compute data centres, and (3) what is the nationality of the accelerator vendors whose chips pow...
Strategic digital sovereignty often takes the form of *techno-blocs*—selective alliances that embed firms in state-led governance of AI infrastructure.
This strategy is rooted in Sovereign Interoperability-a framework that combines selective international openness with strict national control over critical Digital Public Infrastructure (DPI) and Digital Public Goods (DPGs).
By revisiting political theories of sovereignty, examining historical analogies with electricity and the internet, and formalizing the problem as a planner’s allocation under budget constraints, we identified two rules of thumb: equalize marginal returns across the four sovereignty pillars (data, co...
Dimension ... 1. Infrastructure & Connectivity Compute, broadband, cloud access. 2. Data Governance Access, sovereignty, protection. ... • Compute divide: Access to HPC and cloud infrastructure remains the key structural bottleneck, reinforcing vendor lock-in. • Governance gap: Regulatory maturity a...
Sovereign AI is the capability to govern strategically important AI systems in line with national law and public interest, while retaining sufficient operational autonomy over data, compute, model adaptation, and safety controls [18]. Singh and Sengupta propose a four-pillar framework: Data (collect...
My analysis identifies key pathways through which public ownership can shape TAI development, mitigate economic volatility, and distribute AI-derived wealth more equitably.
Sign up to read the full research briefing
Sign up