China AI surveillance drive deepens as ASPI flags mass‑control risks

The ASPI‑based reporting underscores a core tension in the AGI race: the very same advances that make models better at understanding intent, affect and patterns over time also make them ideal instruments for population‑scale surveillance. China’s integration of AI into censorship, predictive policing and so‑called “smart prisons” shows what happens when those capabilities are optimized for regime stability rather than user value. That’s unsettling not just ethically but strategically, because it creates a powerful internal market for sophisticated, continuously‑trained models that are tightly coupled to real‑world decision loops. ([elnacional.cat](https://www.elnacional.cat/es/internacional/china-impulsa-vigilancia-masiva-ia-controlara-ciudadanos-minorias_1529182_102.html))

As Beijing exports this stack – cameras, analytics, behavior‑scoring, and language models for minority languages – it effectively seeds a parallel, state‑aligned AI ecosystem in other autocracies. That could pull talented researchers, data and capital into applications that prioritize control over openness, and it may influence global standards around acceptable AI use. For Western labs arguing that powerful models must remain mostly in private hands, this story is a reminder that geopolitical competitors may go the other way, leaning into state‑run AI infrastructures that treat citizens as training data. That divergence will complicate any future attempt at multilateral guardrails on truly general‑purpose systems.