China develops AI safety benchmark to regulate large models

The Ministry of Industry and Information Technology (MIIT) in China has begun developing a safety standard to evaluate artificial intelligence models, responding to increasing global scrutiny on AI security. The National Industrial Information Security Development Research Centre, under MIIT, is seeking collaboration with companies and experts, with applications closing on Tuesday. The institute highlighted that existing frameworks fall short for complex safety governance, prompting the need for a standardized testing platform for industry compliance. This benchmark will evaluate generative AI on six key aspects: content safety, value alignment, robustness, fairness, privacy protection, and trustworthiness. It will address 31 specific safety risks across five main categories through a hybrid approach, tackling issues like data leaks and hallucinations, in line with international efforts from the US and Europe to strengthen AI oversight.