NB570: Project Glasswing’s FUD and Thunder; Au Revoir Windows, Bonjour Linux

The episode opens with a deep dive into the escalating AI-driven cybersecurity crisis, focusing on Anthropic's announcement of its Frontier model, Mythos Preview, which has reportedly discovered thousands of critical vulnerabilities across major software systems—including every major OS and web browser—and demonstrated the ability to autonomously develop working exploits. This capability has led to the launch of Project Glasswing, a selective initiative granting 40 major tech firms like AWS, Microsoft, Google, and Cisco early access to the model to help secure their systems before public release. The hosts debate whether this is a genuine security effort or a calculated marketing move designed to generate fear, uncertainty, and doubt (FUD), with concerns about insider threats and the risk of privileged information leaking to nation-state actors. They also highlight the growing danger of vulnerability chaining, where low-severity flaws are combined to create high-impact exploits, rendering traditional CVSS scoring inadequate. The discussion shifts to real-world threats, including Iranian cyberattacks on U.S. energy infrastructure targeting Rockwell PLCs, emphasizing the urgent need for air-gapped systems and behavioral threat analytics. Other news includes ARIA Networks' new AI-optimized Ethernet switches, Cisco’s acquisition of AI observability startup Galileo Technologies, Versa Networks achieving FedRAMP medium authorization, Microsoft’s new Azure Service Bus network security perimeter (criticized as a step backward from zero-trust principles), Cato Networks’ zero-time vulnerability protection via inline patching, and Palo Alto’s new cloud firewall migration tool. The episode concludes with France’s push to replace Windows with Linux in government agencies as part of a broader decoupling from U.S. tech, and Amazon’s launch of S3 Files, which brings a file system interface to S3 storage, positioning it directly against traditional NFS servers. The hosts express concern over the speed of exploit development—now under 40 minutes post-patch—and stress the need for automated patching, zero-trust architecture, and proactive security hardening. Key takeaways include: 1) AI is no longer just a tool for security—it’s a weapon capable of autonomous exploit generation, requiring new defensive strategies; 2) Vulnerability management must evolve beyond CVSS scores to account for exploit chaining; 3) Air-gapping critical OT systems is essential for protecting energy and water infrastructure; 4) Zero-trust must be implemented rigorously and continuously, not as a checkbox; 5) The era of 'patch Tuesday' is over—organizations must automate patching and vulnerability remediation; 6) Selective access to powerful AI models like Mythos Preview creates both opportunity and risk, especially for smaller vendors; 7) The move toward European tech sovereignty, while politically motivated, may improve security by reducing reliance on U.S. platforms; 8) AI-driven network optimization and cloud security tools are emerging but require careful evaluation of long-term operational impact and human oversight.