The geopolitical hierarchy of 2028 will depend entirely on who controls the world’s flagship GPUs. According to a new report from Anthropic experts, the U.S. and its allies face a frighteningly narrow window to secure their lead in advanced AI. For now, that edge relies on strict export controls, keeping massive compute—the lifeblood of model training—a scarce resource. However, analysts warn that Chinese labs, under direct state supervision, are closing the gap by exploiting sanction loopholes and siphoning innovation from American research through large-scale distillation attacks.
The outlook for the end of the decade splits into two polar scenarios. In the first, Washington successfully tightens export rules and accelerates deployment, allowing democratic institutions to set the global rules of the game and engage Beijing from a position of strength. The second scenario is a triumph of inertia: authoritarian regimes reach the technological frontier first. Anthropic’s authors argue that the stakes go beyond mere prestige; at risk is the deployment of AI systems for total digital repression and a fundamental shift in the global balance of power in cyberspace.
The race for 'transformative AI' is accelerating, leaving no room for diplomatic niceties. To prevent a future where authoritarian breakthroughs are fueled by American silicon, regulators must act preemptively. Any delay in blocking illegal technology transfers today translates into an irreversible advantage for adversaries tomorrow. Anthropic has essentially issued an ultimatum: either enforce rigorous control over supply chains and architectural secrets, or prepare for a world where global surveillance is automated by an AI that doesn't share your values.