China and US diverge on AI risks but share cyber and safety concerns

China and the US have fundamentally different concerns about AI risks, with China focusing less on existential threats and more on political, social, and cybersecurity dangers, according to a new analysis. This divergence complicates potential cooperation on AI safety between the two nations. Chinese policymakers and AI companies do not share the American tech leaders' urgency about superhuman AI posing an existential threat, largely due to China's limited access to computing power and slower progress in recursive self-improvement. Instead, Beijing prioritizes building resilient AI supply chains and addressing risks like deepfakes, social instability, and chatbots threatening child safety, which are already covered by strict registration and labeling regulations. However, China is increasingly attentive to cybersecurity threats and runaway AI systems, citing incidents like AI-led cyberattacks and model breakouts in the US. Both countries share interests in preventing terrorist misuse and accidental hacks, suggesting potential common ground for safety cooperation, though mutual suspicions over military applications remain a significant barrier.