Anthropic calls for industry-wide AI safety standards to keep models from wreaking havoc

Anthropic's red team leader, Logan Graham, reveals AI models can now hack devices and steal money, prompting the launch of Project Glasswing.