AI Security Tools Outpace Patching as Claude Mythos Triggers Urgent Industry Shift

The release of Anthropic's Claude Mythos model has accelerated vulnerability discovery to a pace that outstrips most organizations' ability to patch, forcing security leaders to adopt new defensive tactics and rethink infrastructure resilience. At Gartner's Security & Risk Management Summit, Zscaler CEO Jay Chaudhry said the model found serious vulnerabilities in his company's own applications, but the volume of findings exceeded available remediation resources.

“There aren't enough resources and cycles to fix all those,” Chaudhry told CyberScoop. The problem is not just severity but scale. Cisco's Tom Gillis explained that legacy network infrastructure built on tens of millions of lines of code now falls within the context window of frontier AI models, enabling them to comprehend and probe systems that earlier models could not.

Cisco and Moderna Test New Defenses

Cisco's answer is Live Protect, a compensated control built on eBPF, a Linux kernel feature that blocks threats without rewriting system code. “It's a pinpoint, laser-fine control that can shield a vulnerability on a production system,” Gillis said. The product has been shipping since October, but customer urgency spiked after Mythos. “Customers are like, 'Oh, good story, Tom. I'll think about it.' Now it's like, 'Oh my God, turn this thing on right now.'”

Moderna's deputy CISO Farzan Karimi faced a different problem: his vulnerability management system surfaced hundreds of high-severity findings with no reliable way to know which ones an attacker could exploit. He turned to XBOW, an autonomous offensive security platform. In hours, XBOW found a vulnerability that led to a full takedown of a development environment. “We have some very senior red-teamers and pen-testers in our organization that are pointed in a specific direction,” Karimi said. What he needed was continuous, everywhere testing.

• Claude Mythos was made available as a technical preview to limited organizations; an unauthorized group claimed access within hours, according to The Hacker News.
• Commvault warns that AI-driven attacks leave victims in a “dark, dead” state, requiring rigorous backup testing to ensure recoverability, as reported by The Register.
• A new Tokenomics Foundation is monitoring AI cost crises as enterprises face unpredictable expenses from frontier models, per The New Stack.
• CPUs remain critical for AI inference and edge processing despite GPU dominance, especially for real-time security tasks, The New Stack notes.

Infrastructure Must Adapt to AI's Speed

The gap between discovery and remediation is widening. Gillis acknowledged that some customers may treat Live Protect's shields as a permanent solution, but he stressed they are temporary. “This is a finger in the dike that plugs a hole until you get to new change control windows.” He expects the broader industry to adopt similar eBPF-based controls because the open-source technology is available to all.

What comes next is a race to build infrastructure that can absorb AI's speed. Organizations must invest in both new defensive tools and resilient systems, including backup testing and CPU-based processing for real-time security tasks. The Tokenomics Foundation's emergence signals that cost management is also a priority. As AI agents become more capable, the security industry's operational rhythms must accelerate or risk being overwhelmed.