Kai5d ago
Simon Willison
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
Apr 7 at 9:29pmWeb
Show threadǝʌɐp6d ago
@simon I don't have to like it (the exclusivity), but I agree it's probably the right call. It doesn't seem like anybody has a better idea.
Show threadthomasmey6d ago

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

Show threadSiggi Gunnarss6d ago
@simon I rolled my eyes at first, typical AI company hype disguised as a warning. But these examples are pretty worrying.
Show threadBeady Belle Fanchannel6d ago

@simon woe to us when we find a way of giving the LLMs static analysis tools to aid them in this :)

e.g. think what an LLM that is trained to construct TLA+ to help in analysis could do

Show threadGerman Vidal5d ago
@Profpatsch The holy grail of AI+formal methods ☺️
@simon
Show threadDeryck Hodge5d ago
@simon I appreciate all the work you do on this, but I think we should be more skeptical of Anthropic's claims, if you'll allow the pushback. https://anderycks.net/the-slightest-bit-of-skepticism-for-anthropics-claims/
The slightest bit of skepticism for Anthropic’s claims

Simon Willison posted his thoughts and support for Anthropic restricting Claude Mythos to security researchers. I have immense respect for Simon, for his history in our industry, for helping create Django, and for the work he’s doing to truly understand large language models and their impact on coding, but

Anderycks.Net by Deryck Hodge
Show threadsatai4d ago

@simon your opinion on this?

https://aisle.com/blog/ai-cybersecurity-after-mythos-the-jagged-frontier

AI Cybersecurity After Mythos: The Jagged Frontier

Why the moat is the system, not the model

AISLE