WASHINGTON (dpa-AFX) - Anthropic has rolled out the Claude Mythos Preview, an unreleased AI model that sits at the heart of its new cybersecurity push, Project Glasswing.
The company claims that Mythos represents a significant advancement in AI, boasting the ability to spot and take advantage of software vulnerabilities on par with top human experts.
Testing revealed that the model identified thousands of serious flaws in major systems think operating systems, web browsers, and popular software. Some of these vulnerabilities had remained hidden for decades, such as a bug in OpenBSD that's been around for 27 years.
Anthropic cautions that Mythos illustrates how AI could significantly reduce the skill and effort needed for cyberattacks, which might ramp up threats to critical infrastructure.
On the flip side, the model's abilities could also be harnessed defensively, allowing for the detection and repair of security gaps faster and on a larger scale than ever before.
Given the associated risks, Anthropic has decided against a public release of Mythos. Instead, they'll be providing access to a select group of partners, aiming to fortify cybersecurity systems while they work on safeguards for future use.
This development marks a pivotal moment where AI might transform both cyber defense approaches and the landscape of cybercrime.
Copyright(c) 2026 RTTNews.com. All Rights Reserved
Copyright RTT News/dpa-AFX
© 2026 AFX News
