Global AI Security Certification Alliance AISecure Established in Geneva: All High-Risk AI Systems Must Pass Third-Party Security Audits

The Global AI Security Certification Alliance AISecure formally established in Geneva on September 10, with founding members including the United States, China, all 27 EU member states, the United Kingdom, Japan, South Korea, and India—28 countries and regions in total. The alliance's core objective is establishing unified AI security audit standards requiring all high-risk AI systems deployed in medical diagnosis, financial decision-making, and judicial assistance to undergo independent third-party security certification.

AISecure's audit framework comprises 127 security test standards covering five dimensions: model robustness, bias detection, privacy protection, explainability, and failure recovery. The alliance will establish 20 certification laboratories worldwide, staffed with trained professional auditors. AI systems passing certification will receive the AISecure mark; systems failing certification will be prohibited from deployment in member states' high-risk domains.

The UN Secretary-General said at the establishment ceremony: "AI technology's development pace has already surpassed any single nation's regulatory capacity. AISecure is our first attempt to establish a global common baseline for AI safety. This isn't about restricting AI development—ensuring that development is safe."

The alliance's first certification work is expected to begin in Q1 2030, targeting completion of 500 high-risk AI system security audits in the first year.