Anthropic's Mythos AI sparks security alarm

The UK's National Cybersecurity Centre has warned that models like Mythos could be used to launch sophisticated attacks on critical infrastructure, including the NHS. Reports have also emerged of a small group of unauthorized users accessing the model through a third-party vendor environment on the same day it was revealed.

Anthropic is currently working with the Pentagon to debunk claims that the company could manipulate the tool once deployed in classified military networks.

The situation has prompted calls in Congress for a new federal framework to manage the risks of high-capability AI systems.