New serious vulnerabilities spiked around release of Claude Mythos Preview

The big number has everyone doing a double take: serious software flaw reports shot up to about 1,500 in June, more than 3.5 times the old monthly high after Anthropic revealed Claude Mythos Preview, an AI system it says can hunt down security holes on its own. Anthropic also says its big-name partners — Microsoft, Google, Apple, and Amazon Web Services — were already using it behind the scenes to find and fix problems before the public ever got near it. Translation for normal people: the robot bug-sniffer may have kicked open a giant closet full of hidden messes.

But the real fireworks are in the comments. One camp is basically popping champagne: "Good" was the blunt mic-drop from one user, while another predicted this is only the appetizer and that way more reports could spill out once the quiet warning period ends. In other words, some readers think this spike is the cleanup phase, not the crisis.

Then came the skeptical squad. One commenter immediately asked the question hanging over the whole story: are all these reports even real, or are some AI-made false alarms? Another threw in the spicier blame game: is AI getting better at finding flaws, or is AI-written code creating more of them in the first place? And of course, the timeline police showed up too, side-eyeing whether people were rushing to connect dots before the release date even matched the graph. The vibe is classic internet: part awe, part panic, part "hold on, the math isn't mathing."