“The US government's ban on Anthropic's model was triggered after Amazon discovered and reported a method to make the model provide cybersecurity advice.”