The AI Safety Game: Beijing’s Conditions for a Productive Summit

The US-China AI talks have become the latest chapter in the ongoing diplomatic standoff between the world’s two largest economies. At its core, this high-stakes game revolves around a fundamental question: what does it mean to set “safety” standards for artificial intelligence, and who gets to define them?

China has refused to engage in bilateral talks on AI until the US proves its domestic developers are playing by the same rules. The Beijing government’s stance is rooted in a scathing critique of Anthropic’s flagship system, Claude, which Chinese commentators accuse of overstepping user privacy thresholds and engaging in secret monitoring.

This may seem like just another episode in the ongoing tech cold war between China and the US. However, Beijing is raising a more nuanced point: who gets to decide what constitutes “safety” in AI development? The Chinese government’s commentary suggests that Washington’s attempts to impose its own corporate guardrails as universal world standards are little more than a thinly veiled attempt to curb China’s digital advancement.

China has long pushed back against Western attempts to dictate the terms of the AI debate. In 2020, Chinese officials accused Google and Facebook of engaging in “digital imperialism” by imposing their own content moderation standards on international platforms. Last year, China passed a sweeping new data protection law that requires foreign companies operating within its borders to comply with local regulations – a move seen as a direct challenge to the US’s efforts to establish global data governance norms.

This highlights the growing tension between competing visions of how AI should be regulated. The US has traditionally taken a more laissez-faire approach, allowing companies like Anthropic to push the boundaries of what’s possible in pursuit of breakthroughs. However, China – and increasingly, other countries – are pushing back against this model, arguing that it prioritizes short-term gains over long-term safety and accountability.

This is not just a technical question; it has deep implications for global governance. As AI becomes increasingly central to national security and economic strategy, the rules governing its development will become ever more crucial. Will we see a multipolar world in which different countries set their own standards, or will some power – whether China, the US, or another country – emerge as the de facto arbiter of AI safety?

The answer is far from clear. But one thing is certain: Beijing’s conditions for a productive summit are not just about setting technical standards; they’re about establishing the terms of the global AI debate.

As tensions between China and the US continue to simmer, it’s worth noting that this is not just a bilateral issue. The Chinese government’s commentary has echoes of a wider global pattern: countries are increasingly pushing back against Western attempts to dictate the terms of AI development. From India’s efforts to develop its own domestic tech sector to the EU’s push for greater regulation of AI, we’re seeing a growing recognition that the future of AI will be shaped by multiple powers – not just Washington and Beijing.

The Chinese government’s critique of Anthropic may be a specific case study, but it highlights a deeper truth about the AI industry as a whole. The US government has long been criticized for its failure to regulate AI development effectively – whether through the lack of strong standards or inadequate enforcement.

As we navigate this complex landscape, one thing is certain: AI safety will become an increasingly pressing concern in the years to come. But who gets to define what constitutes “safety”? Will it be the US government, imposing its own standards on the world? Or will countries like China and India push for a more multipolar approach, one that recognizes the diversity of global perspectives and experiences?

The answer is far from clear – but one thing is certain: the future of AI development will be shaped by competing visions of what constitutes “safety” in this rapidly evolving field.