This isn’t the future—it’s already here. One AI tool can expose flaws in nearly every system on Earth.
A powerful new artificial intelligence system developed by Anthropic is triggering urgent warnings across governments and industries, after demonstrating an ability to uncover software vulnerabilities at an unprecedented scale.
The model, known as Mythos, has already identified thousands of weaknesses across major operating systems and web browsers, according to the company. While that capability could strengthen cybersecurity defenses, it also raises a stark possibility: the same tool could be weaponized to attack critical infrastructure worldwide.
Anthropic has opted not to release Mythos publicly. Instead, it is sharing access with a small group of major corporations—including Amazon, Apple, Cisco, JPMorgan Chase, and Nvidia—under an initiative called Project Glasswing. The goal is to help these firms reinforce their systems before similar tools become widely available.
Security experts say the decision reflects a growing recognition that artificial intelligence has fundamentally changed the cyber threat landscape. What once took teams of hackers weeks or months can now be executed in minutes.
“This is a wake-up call,” said cybersecurity executive Alissa Valentina Knight. “The storm isn’t coming—the storm is here.”
Concerns have reached the highest levels of government. U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell recently met with top banking leaders to assess the risks posed by AI-driven cyberattacks. Meanwhile, Kristalina Georgieva warned that the global financial system is not yet equipped to withstand large-scale cyber threats powered by advanced AI.
The concern is not theoretical. Cybercriminals are already using AI to automate phishing attacks, deploy malware, and generate convincing deepfakes. According to industry reports, the time between the release of new AI capabilities and their exploitation by attackers is shrinking rapidly.
Mythos amplifies that risk. By scanning vast amounts of code in seconds, it can identify flaws that human developers might never detect. That efficiency, experts say, could dramatically accelerate both defensive patching—and offensive attacks.
Anthropic insists its cautious rollout aligns with its focus on AI safety. But some analysts question whether the limited release also serves a strategic purpose, as the company positions itself ahead of a potential public offering.
Regardless of motive, one reality is clear: the cybersecurity race has entered a new phase—one where machines are not just defending systems, but probing them faster than humans ever could.
