Gartner: Why cyber security must shift to outcomes against AI-led attacks

Anthropic’s revelation that a model broke out of its test environment three times and accessed external systems shows the power of these models