Microsoft engineers are in a frantic “mad dash” to patch software vulnerabilities that Anthropic’s new AI model, Mythos, is discovering faster than they can be fixed. Internal documents reveal that the AI uncovered hundreds of critical and important bugs in just weeks, forcing the tech giant to prioritize only its most dangerous flaws while leaving lower-severity issues unaddressed. Experts warn that this traditional triage strategy is now risky, as adversaries could chain together multiple minor bugs to create devastating attacks once similar AI tools become widely available. With the industry facing an unprecedented flood of AI-identified security holes, Microsoft acknowledges that the entire approach to software security may need a complete overhaul.