Crypto
Anthropic releases one-month report card for Mythos, says patch process needs overhaul
Anthropic said its Project Glasswing, involving 50 institutions and companies using the Mythos Preview AI model, found more than 10,000 high-risk or critical software vulnerabilities in a month. In a report, it said security progress now depends on how quickly AI-found flaws are verified, disclosed and patched. It shared results from Cloudflare, Mozilla, Microsoft, Oracle and Palo Alto Networks, and warned that long gaps between discovery, patch creation and deployment increase risk.
