The 'Shanghai Consensus', a collaborative effort signed by over 20 AI experts, notably Geoffrey Hinton and Andrew Yao, highlights that artificial intelligence has attained or even surpassed human intelligence levels, presenting a peril of losing control. This consensus urges the global community to establish clear and enforceable boundaries for AI behavior, ensuring its safety and manageability. It outlines three pivotal actions: conducting comprehensive security assessments, fortifying international collaboration to set universal red lines, and investing in research on 'security by design'. The experts emphasize the necessity of transitioning AI security from reactive defense to proactive construction, fostering the harmonious development of 'benevolence' and 'intelligence', and establishing an international coordination mechanism to implement regulatory and technical standards.