Tencent’s new AI model reflects global tech consolidation amid OpenAI talent exodus and closed-source dominance, deepening corporate control over foundational AI systems

Indigenous knowledge systems emphasize reciprocity, relationality, and non-extractive data practices, starkly contrasting with the proprietary, data-hungry models driving corporate AI development. The closed-source nature of HY3-Preview mirrors colonial patterns of knowledge enclosure, where Western corporations and states control access to critical tools while marginalizing alternative epistemologies. Indigenous AI initiatives, such as those in Māori and Sámi communities, prioritize community ownership and consent, offering a radical alternative to the extractive logics embedded in Tencent’s model.

The migration of OpenAI researchers to Chinese firms like Tencent echoes Cold War-era talent flows, where scientific expertise was weaponized for geopolitical advantage rather than public good. Closed-source AI development mirrors the proprietary software model of the 1980s, which consolidated power in corporations like Microsoft and stifled open innovation. Historical precedents like the Manhattan Project and DARPA’s role in early AI research show how state-military-industrial complexes shape foundational technologies, a pattern now repeating in China’s AI ecosystem.

In Africa, initiatives like Zindi and Deep Learning Indaba prioritize open-source AI for local problem-solving, contrasting with China’s state-backed, closed-source approach. Japanese AI research often emphasizes harmony and societal benefit over competitive innovation, reflecting a cultural ethos that challenges Western meritocratic narratives. Latin American countries like Brazil and Mexico are exploring AI governance models that integrate indigenous cosmovisions, such as Buen Vivir, to counter extractive data practices.

The closed-source nature of HY3-Preview limits reproducibility and peer review, undermining scientific rigor and slowing collective progress in AI safety research. The model’s relatively small size (295B parameters) may reflect efficiency gains from proprietary optimization, but without transparency, it’s unclear how it compares to open models like Llama 3 or Mistral. Scientific consensus increasingly favors open models for their ability to democratize access and enable independent audits, a principle ignored in Tencent’s approach.

Artistic and spiritual traditions often frame AI as a mirror of human consciousness, raising questions about the ethics of creating models that replicate human biases without moral frameworks. In Buddhist thought, the concept of 'dependent origination' critiques the illusion of isolated innovation, highlighting how Tencent’s model is embedded in a web of corporate, state, and global power structures. Indigenous storytelling traditions warn against the hubris of creating 'autonomous' systems, emphasizing instead relational intelligence and interdependence.

A future dominated by closed-source AI models risks entrenching corporate monopolies, reducing innovation to proprietary silos, and exacerbating global inequalities in access to critical technologies. Scenario modeling suggests that if China and the US continue to prioritize state-backed, closed-source AI, the rest of the world may become dependent on their infrastructure, limiting sovereignty in AI governance. Conversely, open-source models could enable decentralized innovation, but only if accompanied by robust regulatory frameworks to prevent misuse.

Marginalized researchers, particularly those outside elite institutions in the Global South, are systematically excluded from the narrative of AI progress, despite contributing critical insights to ethical AI design. The exodus of OpenAI researchers to Tencent reflects a brain drain from public-interest AI research, leaving underfunded institutions in the West and Global South further behind. Indigenous communities, who bear the brunt of data extraction and surveillance, are rarely consulted on AI development priorities, despite their unique perspectives on consent and reciprocity.