China’s 6th National AI Competition forum spotlights model battles and publishes new AI safety benchmarks and evaluation results

Summary

A forum tied to the 6th China AI Competition in Xiamen highlighted a shift in China’s AI ecosystem from “bigger models” toward measurable capabilities and safety governance, with dedicated tracks including large-model adversarial contests and AIGC video detection. Organizers said the competition drew nearly 700 teams and around 1,200 submissions—useful as a barometer of where academic and industrial talent is clustering (and which tasks are becoming standardized). The accompanying AI safety governance forum reported the release of AI security benchmarks and a new batch of AI security evaluation results, pushing the industry toward compliance-style scoring rather than purely benchmark-chasing. Notably, representatives from Huawei, iFlytek, Alibaba Cloud, Baidu, China Telecom’s AI research arm, and QiAnXin were referenced as part of the broader expert/enterprise participation—signaling that security vendors and telecoms are increasingly central to China’s “trust layer” for deployed AI. The takeaway for the Race-to-AGI landscape: national contests are being used not just to discover talent, but to align industrial roadmaps around safety baselines, detection tooling, and operational requirements for real-world deployment.