“The UK's AI Security Institute (AISI) intends to test the Kimi K3 model on its cybersecurity benchmarks once the model's weights are publicly released.”