Chinese AI firms shift to proprietary models amid global open-source tensions, revealing structural shifts in tech sovereignty and revenue extraction

Indigenous knowledge systems emphasize collective ownership of knowledge, contrasting sharply with the proprietary enclosure driving Chinese AI firms. Many Indigenous communities view AI as a tool for cultural preservation rather than revenue extraction, as seen in projects like the Māori-developed Te Hiku Media’s AI for language revitalization. The shift to proprietary models risks erasing these alternative frameworks, replacing them with extractive logics that prioritize profit over communal benefit.

The pivot to proprietary models echoes historical tech enclosures, such as the Unix wars of the 1980s or IBM’s proprietary shifts in the 1970s, where open collaboration gave way to corporate control. China’s AI industry has long relied on state-led open-source initiatives (e.g., PaddlePaddle) to build domestic capacity, but U.S. sanctions have forced a retreat into proprietary models. This mirrors Cold War-era tech nationalism, where geopolitical tensions accelerated the fragmentation of global tech ecosystems.

Cross-culturally, the tension between open and proprietary models reflects differing values around knowledge: Western corporate models prioritize exclusivity, while African and Indigenous traditions emphasize communal access. In China, the state’s role in directing AI development contrasts with Silicon Valley’s libertarian ethos, yet both systems converge on extractive models. The rise of proprietary AI in China also parallels India’s push for indigenous AI stacks, highlighting a global divide between corporate and public-interest approaches.

Scientifically, the shift to proprietary models is driven by the computational and economic infeasibility of open-sourcing large models, which require massive energy and hardware investments. Research shows that open-source models often lag behind proprietary ones in performance due to limited access to training data and compute. However, open-source models democratize access, enabling innovation in low-resource settings—a critical factor for global AI equity.

Artistically, the proprietary model trend risks reducing AI to a tool of capital accumulation, stripping it of its potential as a medium for creative expression and cultural dialogue. Indigenous and spiritual traditions often frame technology as a sacred trust, not a commodity, as seen in the Māori principle of 'kaitiakitanga' (guardianship). The enclosure of AI models may stifle artistic collaboration, where open-source tools have enabled projects like AI-generated poetry or music that transcend corporate boundaries.

Future modelling suggests that the proprietary shift could lead to a fragmented AI ecosystem, where access is gated by corporate or state power, exacerbating global inequalities. Scenario planning indicates that without intervention, 80% of AI innovation may become concentrated in a handful of firms by 2030, with open-source alternatives relegated to niche roles. However, public-interest AI models (e.g., EU’s open-source initiatives) could counter this trend by prioritizing accessibility and transparency.

Marginalized voices include Chinese researchers in academia who rely on open-source tools but face increasing restrictions due to corporate and state policies. Global South developers are excluded from proprietary ecosystems, limiting their ability to innovate in local languages or contexts. Women and non-binary technologists, who are underrepresented in AI development, are further sidelined as proprietary models centralize power in the hands of a few elite firms.