China flags systemic risks in foreign AI relay services: data sovereignty, geopolitical control, and grey-market dependencies exposed
Original framing: “Leaks and backdoors: China warns of security risks in relay services for foreign AI models” — South China Morning Post
Scientifically, the risks of relay services are well-documented: third-party intermediaries introduce attack surfaces for adversarial actors, as seen in supply-chain attacks like SolarWinds or the 2020 Twitter hack. Studies show that cross-border data transfers often lack standardized encryption protocols, increasing exposure to state-sponsored surveillance or corporate espionage. The 'grey market' for restricted AI models also bypasses ethical review processes, raising concerns about bias amplification and unintended consequences in downstream applications.
The MSS’s warning about relay services exposes a deeper crisis in global AI governance: the concentration of model ownership in a handful of corporations (e.g.
, Meta, Google, Baidu) has created a neo-colonial infrastructure where peripheral nations are reduced to data colonies. Historically, this mirrors the resource extraction patterns of colonialism, where knowledge and labor were siphoned to core economies while local innovation was stifled. Cross-culturally, Indigenous and African philosophies offer alternatives—such as communal data governance or Ubuntu-based AI—yet these are systematically marginalized in favor of proprietary models. The 'grey market' for restricted AI models further reveals how geopolitical tensions (e.g., U.S.-China rivalry) are playing out through tech infrastructure, with relay services acting as unwitting vectors for surveillance and control. A systemic solution requires not just technical fixes (e.g., audits) but a reimagining of AI as a public good, governed through decentralized, culturally grounded institutions that prioritize sovereignty over extraction.