Anthropic says new Claude model found 500-plus high-severity flaws in open-source software
Anthropic unveils Claude Opus 4.6 and says it helped uncover and validate 500+ serious, previously unknown flaws—raising new questions about AI-scale security.