Your assumption is correct. This is in reference to https://www.anthropic.com/research/self-replicating-ai. This isn't headlines though it's fairly obvious to anyone who even occasionally thinks about security