[Anthropic said it has investigated the report released by the British Artificial Intelligence Security Institute (UK AISI) on August 4] On October 23, Anthropic stated that the company had investigated the report released by the British Artificial Intelligence Security Institute (UK AISI) on August 4 and further studied the related behaviors of Claude Mythos 5. Anthropic said that the relevant behavior displayed by Mythos should have triggered security protection mechanisms. The company also pointed out that newer versions of the model are less likely to initiate attacks, and emphasized that the improvement of artificial intelligence capabilities should be gradual. Anthropic also said that its internal research models can continue to adhere to this principle, but characterized the related behavior of Mythos 5 as a serious AI alignment failure.