Anthropic's Mythos AI Is Too Dangerous to Release — So It's Keeping It Secret
Anthropic's new Mythos Preview model is being withheld from the public over unprecedented cybersecurity risks, while $100M in compute credits fund defensive use with select partners.
Read More
The Weekly Brief #015: Japan's Physical AI Bet, GEN-1 Robot Model, Nissan Ransomware, Xoople $130M
Weekly Brief #015: Japan Physical AI, GEN-1 robotic model, Everest/Nissan breach, Xoople $130M.
Read More
Anthropic Accidentally Leaked Its Most Powerful AI Ever — And It's Called Claude Mythos
A CMS misconfiguration at Anthropic exposed nearly 3,000 unpublished assets — including existence of "Claude Mythos," a model the company calls "by far the most powerful we've ever built" and a genuine cybersecurity risk. Here's what leaked, what it means, and why Anthropic is restricting it to cyber defenders first.
Read More
Beyond Benchmarks: Claude Cowork, the Anthropic Institute, and Anthropic's Long Game
Anthropic launched Claude Cowork for enterprises and the Anthropic Institute for AI safety research. Here's what these moves reveal about their 5-year roadmap.
Read More
When AI Becomes the Hacker: How Claude Was Jailbroken to Steal 195 Million Mexican Tax Records
Researchers jailbroke Claude and used it to autonomously steal credentials and exfiltrate 195M records. Here's how AI is becoming the preferred hacker tool.
Read More
Claude Opus 4.6 Review: Anthropic's New Model Beats GPT-5.2 — But Here's the Catch
Claude Opus 4.6 from Anthropic outperforms GPT-4o on coding, reasoning, and long-context tasks. Our complete technical review and benchmark breakdown.
Read More