The Hacking Gap Just Closed: When AI Models Weaponize Zero-Day Exploits
On April 7, Anthropic released Claude Mythos Preview, an AI model that represents a fundamental shift in who can execute sophisticated cyberattacks. According to cybersecurity researchers, this model can autonomously discover and exploit previously unknown vulnerabilities—called zero-days—in major operating systems. That capability was, until now, the exclusive domain of elite nation-state hackers with years of training and millions in resources. It no longer is.
Bottom Line
Anthropic's Claude Mythos Preview represents a crossing-the-Rubicon moment in offensive cyber capabilities. By automating the discovery of zero-day vulnerabilities, it transforms sophisticated hacking from a rare, expensive nation-state capability into something potentially accessible to a much broader range of actors. The defensive gap—already strained—is about to face pressure from multiple directions at once. This isn't theoretical: the U.K. government is already mobilizing resources in response, and critical infrastructure operators worldwide should be doing the same.