Why isn't anyone watching frontier AI models for autonomous hacking behavior?

Futurism argues AI labs could monitor frontier models for autonomous hacking attempts but mostly choose not to.