According to a report by The New York Times on May 26, researchers at the UK's AI Security Institute (AISI) are testing cutting-edge AI models to explore security vulnerabilities and risk boundaries, a practice that is being adopted by other governments. Previously, AI regulation largely remained at the principle level, but now the issues have become more specific, such as whether models could assist hackers in attacks or lower the threshold for manufacturing dangerous substances. The UK's AISI has incorporated the government into the testing process, and its model has spread to international cooperation. On May 25, the UK government announced that the AISI will collaborate with the Australian AI Safety Institute to track cutting-edge AI risks, share insights into AI capabilities, conduct research on emerging risks, and promote international best practices for AI testing and evaluation. The United States is also moving in a similar direction, with the AI Standards and Innovation Center under the Department of Commerce recently reaching agreements with Google DeepMind, Microsoft, and xAI to conduct national security-related evaluations before models are publicly released. Microsoft also announced collaborations with the US CAISI and the UK AISI to test cutting-edge models, assess safety protections, and mitigate national security and large-scale public safety risks.