The Second Global Conference on Artificial Intelligence, Security, and Ethics (AISE26) was held in Geneva from June 18 to 19, 2026, focusing on the impact of artificial intelligence on international peace and security. The core consensus of the conference was that AI security assessment conclusions must be incorporated into decision-making processes to constrain system behavior, rather than merely serving as procedural information. Discussions covered three dimensions: promoting model testing towards system security, ensuring that assessments truly influence deployment decisions, and enhancing the judgment capabilities of more countries. It was emphasized that security requirements should change with usage and consequences, agents should have limited permissions and review mechanisms, and human supervision must be substantially effective. Responsibilities should be clearly defined along the development, integration, and deployment processes, and transparency design should support verifiable information. China has issued the “Measures for Identifying Content Generated by Artificial Intelligence,” implementing identification obligations at all stages of the dissemination process. International cooperation aims to bridge the intelligence gap, turning principled commitments into practical constraints in system operation through capacity building.
The Second Global Conference on Artificial Intelligence, Security, and Ethics (AISE26) was held at the Palais des Nations in Geneva from June 18 to 19, 2026, focusing on the impact of artificial intelligence on international peace and security. The conference emphasized that evaluation conclusions must be incorporated into decision-making processes to constrain system behavior, rather than merely serving as procedural information. Discussions covered three dimensions: moving from model testing to system security, ensuring that evaluations truly influence deployment decisions, and enabling more countries to make judgments. It was agreed that security requirements should change with the purpose and consequences of use, agents should have limited permissions and review mechanisms should be established, and human supervision must be effectively implemented. Responsibilities should be clearly defined along the development, integration, and deployment processes, and transparency design should support verifiable information. China has issued the “Measures for Identifying Content Generated by Artificial Intelligence,” implementing identification obligations at all stages of dissemination processes. International cooperation aims to bridge the intelligence gap, making principle commitments into actual constraints in system operation through capacity building.