Link to: https://blog.cryptographyengineering.com/2026/07/29/some-notes-about-anthropics-new-results/
Matthew Green reports measurable progress in Anthropic's models on targeted cryptographic problems over the past five months, validating AI capability beyond generic autocompletion. For tech leaders, this reinforces that frontier models excel in specialized domains but exhibit sharp performance boundaries—a crucial constraint in evaluating deployment readiness and domain fit.
Matthew Green on Anthropic’s New Cryptanalysis Results
Matthew Green:
If you’re under the impression that these models are “glorified
autocomplete” or that progress is slowing down, I need to urge
you: stop thinking that. The models are very intelligent and
Yesterday Anthropic published two new cryptanalysis results, both outputs of Claude Mythos, their (still) unreleased advanced…
The company says it has built its most dangerous model yet. Can its coalition of internet companies fix the internet before…
Anthropic's Claude Mythos Preview autonomously exploits zero-day vulnerabilities. Access is limited to Project Glasswing, a $100M…
Anthropic said it was testing the new model, which it called a “step change” in performance, after accidental data leak reveals…
Anthropic says its Claude Mythos Preview AI model discovered improved mathematical attacks against the HAWK post-quantum digital…
: Emil Michael says agencies are evaluating the cybersecurity model, not deploying it
