DeepSeek’s Disruption: Digital “Sovereignty” Meets Reality at the Paris AI Summit

As policymakers and tech luminaries descend on Paris for next week’s AI Action Summit, they are already finding their carefully crafted agenda upended by an unexpected gatecrash: DeepSeek.

The Chinese firm claims to have developed a large language model rivaling the best American offerings, including OpenAI’s ChatGPT, at a fraction of the usual cost. Chinese AI innovation has not merely disrupted the Summit’s choreography—it has called into question fundamental assumptions about digital sovereignty: the latest geopolitical battlefield where nations compete for control over AI capabilities, data flows, and technological infrastructure just as they once fought over physical trade routes.

The timing could hardly be more pointed. As French President Emmanuel Macron prepares to champion what he calls “AI biodiversity,” encouraging the development of these emerging technologies by more stakeholders like France’s homegrown Mistral AI, DeepSeek’s announcement shows that significant AI innovation can emerge unexpectedly. The industry pivot comes as Summit envoy Anne Bouverot warns against AI being “captured by a few predators.”

This vision aligns with the recently passed EU AI Act, which includes “diversity, non-discrimination, and fairness” among its core principles and seeks to promote an AI that “includes diverse actors and promotes equal access.” Yet as Senator Josh Hawley’s proposed legislation to criminalize DeepSeek usage in the United States demonstrates, some American policymakers are responding to Chinese AI advancement not with competitive innovation, but with digital isolationism that could fragment the global AI ecosystem even further.

Now, DeepSeek appears to have demonstrated that the technology’s commanding heights may be more accessible than previously imagined. The Chinese company claims its AI cost just under $6 million to build—loose change compared to the hundreds of millions that American firms are pouring into AI infrastructure.

More remarkably still, DeepSeek appears to have achieved this feat while operating under American semiconductor sanctions—efforts to curb Chinese innovation by limiting the export of the most sophisticated chips. Where Meta’s latest model required 16,000 chips to train, DeepSeek claims to have managed with just 2,000. If true, this represents more than mere technical ingenuity—it suggests that attempts to maintain technological supremacy through export controls may be fundamentally misguided.

The implications for the whole concept of “digital sovereignty” are profound, if paradoxical. At first blush, DeepSeek’s achievement might appear to vindicate China’s pursuit of technological self-sufficiency. Yet it actually demonstrates the futility of trying to erect national moats around AI development. America’s attempt at containment through chip restrictions is perhaps the most ambitious experiment in digital sovereignty to date. But while the U.S. seeks to limit the global spread of AI by controlling the outflow of technology, it may have inadvertently accelerated Chinese innovation.

The DeepSeek breakthrough challenges established narratives about AI development costs. Just last year, NVIDIA’s Jensen Huang toured globally promoting “sovereign AI,” assuring leaders it was “not that costly” while positioning his company’s expensive GPUs as essential infrastructure. DeepSeek’s achievement suggests a different path: one that relies less on premium hardware and more on innovative engineering.

“Without a voice in shaping AI governance, developing countries risk being trapped in a cycle of technological dependence, forced to adopt tools and policies ill-suited to their local circumstances.”

Despite the corporate co-option of “sovereign” language, digital sovereignty might not be the best path to a more equitable and inclusive digital future. Instead, there’s a more nuanced approach that the AI Summit’s architects would do well to consider: digital agency. Digital agency centers the rights and needs of communities and individuals within nations. At its core, the concept is about recognizing the multitude of stakeholders at all levels of governance that make up our global digital ecosystem; harnessing the positive potential of technology; and promoting collaboration. For policymakers, this translates into three concrete priorities: (1) reinventing governance institutions, (2) enabling “edge technologies” like mobile phones for remote health care or decentralized storage networks for more secure data management, and (3) building human capacity alongside technical capability.

The stakes could not be higher, particularly for the Global South. Most AI policies and practices have been developed by and for wealthy nations, creating what a recent New America and Igarapé Institute brief calls an “AI governance divide.” Without a voice in shaping AI governance, developing countries risk being trapped in a cycle of technological dependence, forced to adopt tools and policies ill-suited to their local circumstances—or to embrace corporate visions of sovereignty that may serve shareholders more than citizens.

The G20, under South Africa’s presidency, could help bridge this divide through an equitable AI development forum. But the real need may be more fundamental: a multistakeholder governance system that accounts for both regional capabilities and the intensive resource demands of AI development—from energy and minerals to data and human capital.

DeepSeek’s emergence offers a timely illustration of these dynamics. If its approach proves replicable, it could democratize access to advanced AI capabilities far beyond today’s tech titans. Yet questions abound: OpenAI has accused DeepSeek of using its proprietary models, while European regulators have blocked or scrutinized the chatbot over privacy concerns. Even if its technical claims prove true, computational efficiency alone cannot solve fundamental challenges like data quality and availability.

The consequences of getting this wrong would not respect national boundaries. The ripple effects of an AI governance divide would be felt globally, from labor displacement to widening inequality. Innovation transcends borders and corporate strategies. For Summit attendees, the real task is not to enforce technological barriers but to ensure AI’s benefits and risks are shared equitably across an increasingly interconnected world.

The Summit may not resolve these tensions. But if it can help shift the conversation from national competition and corporate interests toward collaborative governance, particularly by amplifying Global South voices, it may prove to be a crucial milestone in the global AI journey. After all, as DeepSeek demonstrates, sometimes the most significant breakthroughs come from unexpected places.

Follow The Thread! Subscribe to The Thread monthly newsletter to get the latest in policy, equity, and culture in your inbox.