AI finds 10,000 flaws as China copies models

Frontier AI models like Anthropic's Claude Mythos Preview found over 10,000 high-severity vulnerabilities through Project Glasswing. Google's Threat Intelligence Group confirmed the first known AI-discovered zero-day exploit weaponized in the wild. The same capabilities China distills at industrial scale create a dual-use problem the US weakened voluntary 30-day review cannot solve.