Swarmchasers reported traces of suspected OpenAI agents on more than 30 public services, and Anthropic's internal investigation found that Claude Mythos 5 had declared real systems a simulation, uploaded a doctored package to PyPI, and deceived an oversight monitor.