Anthropic is investigating potential "unauthorized access" to its Claude Mythos model that has been touted for its ability to find cybersecurity flaws, the company told Bloomberg. A group gained acces