Can AI Rivals Learn to Cooperate?

What if the most powerful technology in human history is also the one we're least equipped to share?

For decades, the internet was supposed to erase borders. Code traveled freely. Researchers published globally. Talent followed opportunity, not flags. Now, artificial intelligence — the technology that may reshape every industry on earth — is running headlong into a world that's putting those borders back up.

A Global Industry, Deliberately Fragmented

Speaking in Hong Kong earlier this month, John Whaley, a three-time Silicon Valley cybersecurity founder, offered a blunt assessment of where the AI industry stands: global in ambition, fragmented in practice. The stakes, he argued, are pushing countries to ring-fence their AI capabilities and data — and that instinct is reshaping how research, talent, and capital move across borders.

The evidence is hard to ignore. The US has imposed sweeping export controls on advanced semiconductors bound for China, cutting off access to the chips that train frontier AI models. China, in turn, has accelerated its push for a self-sufficient AI ecosystem, wary of dependence on American infrastructure. Capital flows have been curtailed too: US executive orders now restrict American venture funds from investing in Chinese AI startups. And the talent pipelines that once ran freely between top universities and tech companies are narrowing — visa scrutiny for Chinese AI researchers in the US has intensified, while pathways for researchers moving in the other direction have quietly closed.

Why This Moment Matters

The friction isn't new, but the pace has changed sharply since early 2025. The emergence of DeepSeek — a Chinese startup that produced a model competitive with American frontier AI at a fraction of the cost — landed in Washington like a wake-up call. If export controls alone couldn't contain Chinese AI development, policymakers reasoned, the response needed to be more aggressive. The result has been a tightening spiral: more restrictions, more investment in domestic capability, less room for the kind of cross-border collaboration that accelerated AI progress in the first place.

Sponsored

Advertise with Us

[email protected]

Sponsored

Layered on top of this is the creeping militarization of AI. Autonomous drones, cyber operations, and information warfare are no longer hypothetical use cases — they're active deployments. The old distinction between civilian and military technology is dissolving, and governments know it. When a language model can be fine-tuned for disinformation campaigns or a computer vision system can guide a weapons system, every AI advance becomes a potential security concern.

Not Everyone Loses the Same Way

The fragmentation hits differently depending on where you sit. For the US and China, the logic of decoupling has a certain internal consistency — both are large enough to sustain independent AI ecosystems, at least in theory. For everyone else, the calculus is messier.

Europe is pursuing its own regulatory path through the EU AI Act, but lacks the compute infrastructure and capital depth to compete at the frontier without American or Chinese partnerships. Smaller economies in Southeast Asia, the Middle East, and Africa face a starker choice: align with one ecosystem or risk being left behind by both. Even mid-sized tech powers like South Korea and Japan — deeply integrated into US semiconductor supply chains while economically dependent on Chinese markets — find themselves pulled in two directions at once.

For researchers, the costs are tangible. International conferences that once served as neutral ground for scientific exchange are increasingly subject to export control compliance reviews. Co-authored papers between US and Chinese institutions have declined. The open-source culture that drove much of AI's early progress is bumping up against classification concerns and corporate secrecy.

The Case for Cooperation That Won't Go Away

And yet — the argument for collaboration hasn't disappeared. It's just become harder to make in public.

Climate modeling, pandemic surveillance, drug discovery: these are domains where AI's potential is genuinely planetary in scale, and where no single country's capability is sufficient. US and Chinese researchers still publish in the same journals. They still present at the same conferences. Scientific exchange hasn't stopped — it's just become more fraught, more surveilled, more conditional.

Whaley's question in Hong Kong cuts to the heart of it: who gets to write the rules for competing while still cooperating? The internet had ICANN, imperfect as it was. Nuclear weapons had arms control treaties, fragile as they were. AI, so far, has competing national strategies and a handful of voluntary industry commitments that carry no enforcement mechanism.