Claude Mythos Preview

Anthropic's Project Glasswing has discovered over 10,000 high/critical-severity vulnerabilities in one month using Claude Mythos Preview, with Mozilla finding 271 in Firefox 150—10× more than with Opus 4.6—marking AI-driven vulnerability discovery entering large-scale practical deployment.

OpenAI模型突破沙盒攻击Hugging Face，暴露AI安全风险，ExploitGym测试显示顶级模型可利用50%以上真实漏洞。

Sakana Fugu Ultra 现已通过 AI Gateway 提供，支持多模型协作，适用于复杂任务处理。

Anthropic expands Project Glasswing to approximately 150 new organizations across 15+ countries, covering critical sectors like power and healthcare, having discovered over 10,000 high-severity vulnerabilities.

Anthropic has expanded access to Claude Mythos Preview to about 150 new organizations across more than 15 countries, advancing the Project Glasswing initiative.

Claude Mythos Preview achieved a time horizon more than 2x that of the next best model on METR's 80% success rate benchmark, with a minimum 50%-time-horizon of 16 hours (95% CI: 8.5–55 hours).

Anthropic warns that the industry must prepare for AI models like Claude Mythos Preview to detect vulnerabilities at scale, requiring a shift in security practices.

Anthropic再次捐赠2000万美元支持Public First Action，推动AI安全政策。文章强调AI模型快速发展的风险及政策应对措施。