Anthropic AI Uncovers Security Flaws in Major OS and Browsers

A new AI-driven cybersecurity initiative is launching through a collaboration between Anthropic and several major technology firms, including Nvidia, Google, Amazon Web Services, Apple, and Microsoft. This partnership, known as Project Glasswing, aims to provide large organizations and government entities with a powerful tool for autonomously identifying critical security weaknesses in their software and infrastructure. The core of the project is a specialized AI model designed to scan for vulnerabilities with minimal need for human oversight.

As part of this effort, Anthropic is providing its initial partners with exclusive access to a new model called Claude Mythos Preview. This is a general-purpose AI system that the company has decided not to release publicly at this time, citing significant security considerations. According to Newton Cheng, who leads the frontier red team for cybersecurity at Anthropic, the model is intended to empower security professionals by automating the complex process of discovering flaws. The goal is to enhance defensive capabilities by proactively finding issues before malicious actors can exploit them.

The initiative has already demonstrated its potential by uncovering previously unknown security vulnerabilities within widely used operating systems and web browsers. This early success highlights how advanced AI models can be leveraged to strengthen digital defenses across the industry. By integrating this technology, companies hope to shift from reactive security patches to a more proactive and continuous assessment of their systems’ integrity.