Anthropic and OpenAI warn Claude Mythos breaks into corporate networks

The author argues that falling AI compute costs have dramatically shifted the economics of corporate network security in favor of attackers. While frontier models like Claude Mythos and Zhipu's GLM-5.3 can now execute automated corporate network intrusions for as little as ninety dollars, security operations centers face escalating defense costs compounded by speed, scale, and a massive surge in non-human identities. Open article →