Wrote up some thoughts on Anthropic's Project Glasswing, where their latest Opus-beating model is available to partnered security research organizations only. Given the recent alarm bells raised by credible security voices I think this is a justified decision.
https://simonwillison.net/2026/Apr/7/project-glasswing/
Anthropic’s Project Glasswing—restricting Claude Mythos to security researchers—sounds necessary to me

Anthropic didn’t release their latest model, Claude Mythos (system card PDF), today. They have instead made it available to a very restricted set of preview partners under their newly announced …

Simon Willison’s Weblog

RE: https://fedi.simonwillison.net/@simon/116365550190475673

finding those kind of error is really worrying, as it demonstrates a deep understanding of how things play together.

The potential flood of findings OTOH is probably already too high as tools like syzcaller already find so many potential bugs:
https://syzkaller.appspot.com/upstream

#deepblue