AI Model Mythos Finds Thousands of Old Security Flaws

Thousands of dangerous software bugs that cybercriminals could have exploited for decades are finally being found and fixed, thanks to a powerful new AI system called Mythos.

AI company Anthropic just launched Project Glasswing, a security initiative where 12 major tech partners including Apple, Microsoft, Amazon, and Cisco are using the advanced AI model to scan software for hidden weaknesses. In just a few weeks, Mythos discovered thousands of previously unknown vulnerabilities, many considered critical security risks.

What makes this especially impressive is that many of these bugs are one to two decades old. They've been sitting undetected in both proprietary and open source software that millions of people use every day.

Mythos wasn't specifically designed for cybersecurity work. It's actually a general-purpose AI with strong coding and reasoning abilities. But those skills turned out to be exactly what's needed to spot the kinds of complex security flaws that human reviewers might miss.

The 12 partner organizations include some of the biggest names in tech and cybersecurity: Amazon, Apple, Broadcom, Cisco, CrowdStrike, the Linux Foundation, Microsoft, and Palo Alto Networks. Together, they're scanning both their own software and critical open source systems that power much of the internet.

The best part? These companies have committed to sharing what they learn with the broader tech industry. When one partner discovers a vulnerability and figures out how to fix it, that knowledge will help everyone build more secure software.

Beyond the initial 12 partners, 40 additional organizations will get access to test Mythos for security work. While the model won't be available to the general public, this collaborative approach means its benefits will spread far beyond those directly using it.

The Ripple Effect

This initiative represents a fundamental shift in how the tech industry approaches cybersecurity. Instead of each company protecting only its own products, major competitors are working together to make the entire digital ecosystem safer.

Every vulnerability found and patched is one less opening for hackers to steal data, disrupt services, or compromise personal information. Given that some of these flaws have existed for 20 years, fixing them could prevent security breaches that might have affected millions of people.

The collaborative nature of Project Glasswing also sets a powerful precedent. When tech giants share security knowledge instead of hoarding it, everyone from small startups to individual users benefits from stronger, more trustworthy software.

As AI continues to advance, using these powerful tools to defend rather than exploit could help tip the scales in favor of the good guys in the ongoing battle for digital security.