“Kate Judge of Columbia University believes advanced AI systems like Anthropic's Mythos will likely detect significant vulnerabilities in banking software, posing a systemic risk of disabling cyberattacks.”