Anthropic is expanding its restricted cybersecurity AI program to roughly 200 partners

The company said partners using its Claude Mythos Preview model have already found more than 10,000 high or critical security flaws