- Anthropic released a report titled “2028: Two scenarios for global AI leadership,” outlining two scenarios of AI competition between the US and China by 2028.
- The report suggests that AI will soon be powerful enough to shift the global balance of power, support mass citizen surveillance, and provide superior military advantages.
- Anthropic asserts that the US’s greatest advantage currently lies in compute — advanced AI chips controlled by US companies and their allies.
- According to the report, Chinese AI labs are still closely trailing the US by leveraging talent, exploiting chip export control loopholes, and executing “distillation attacks” to clone US AI models.
- Anthropic forecasts that if the US tightens chip controls, prevents distillation, and boosts the export of Western AI, democracies can maintain a 12–24 month technological gap through 2028.
- The report warns that if the US relaxes compute controls, China could catch up with or surpass the AI frontier and shape global AI standards according to a state-controlled model.
- Anthropic is particularly concerned that AI could help the Chinese government expand surveillance, censorship, and automated repression on an unprecedented scale.
- The company emphasizes that frontier AI is also a dual-use technology, capable of being used for cybersecurity, drone coordination, vulnerability detection, and military combat support.
- The report cites an example where Claude Mythos Preview helped Firefox fix more security bugs in a single month than in the entirety of 2025, and nearly 20 times the previous monthly average.
- Anthropic believes the AI race is currently unfolding across four fronts: model capability, domestic deployment, global distribution, and the ability to maintain social stability.
- According to the analysis, Huawei will only generate about 4% of NVIDIA’s compute capability in 2026, dropping to 2% in 2027.
- The report calls on the US to tighten chip exports, close foreign data center loopholes, combat GPU smuggling, and restrict access to US AI models to prevent China from narrowing the gap.
- 📌 Conclusion: Anthropic views the US–China AI competition as the most critical geopolitical and national security issue of the coming decade. The report notes that compute is the deciding factor for global AI dominance, and the US currently maintains a significant lead thanks to its advanced chip ecosystem. However, Anthropic warns that this gap could rapidly disappear if China continues to exploit chip export loopholes and distill US AI models.
