AWS and the New Era of AI-Driven Cybersecurity

Securing the world’s most critical digital infrastructure requires more than just traditional firewalls; it demands a proactive, AI-driven strategy that evolves faster than the threats it faces. Under the leadership of Vice President and Chief Information Security Officer (CISO) Amy Herzog, Amazon Web Services (AWS) is redefining cloud security by integrating advanced AI to hunt for vulnerabilities before they can be exploited.

Scaling Defense: The Magnitude of AWS Security

The sheer scale of AWS operations necessitates an automated approach to threat detection. To protect millions of customers globally, AWS security and threat intelligence teams analyze over 400 trillion network flows every single day to identify patterns that signal emerging threats. This massive data processing capability allows AWS to strengthen defenses for all customers based on the threats detected for a single user.

The effectiveness of this proactive stance is evident in recent statistics. In 2025 alone, AWS blocked over 300 million attempts to maliciously encrypt customer files hosted on Amazon S3.

Slashing Response Times with AI Automation

One of the most significant bottlenecks in cybersecurity is the time security operations (SecOps) engineers spend analyzing logs. AWS has addressed this by implementing an AI-powered log analysis system that has fundamentally changed the speed of response.

This system has reduced the average time spent analyzing security logs from six hours to just seven minutes. This 50x productivity increase allows engineers to detect and respond to threats with unprecedented speed, minimizing the window of opportunity for attackers.

Project Glasswing and Claude Mythos Preview

AWS is playing an integral role in advancing a new class of AI specifically designed for cybersecurity through a partnership with Anthropic. This initiative, known as Project Glasswing, aims to secure the world’s most critical software and advance the industry’s cybersecurity practices.

What is Claude Mythos Preview?

At the heart of Project Glasswing is Claude Mythos Preview, Anthropic’s most advanced AI model to date. Unlike previous frontier models, Claude Mythos Preview represents a step-change in reasoning and capabilities, specifically optimized for:

• Cybersecurity tasks: Identifying and patching vulnerabilities in critical digital infrastructure.
• Software coding: Writing more secure code and analyzing complex software architectures.
• Complex reasoning: Solving intricate security problems that require deep logical analysis.

A Proactive Security Philosophy

The core philosophy driving these advancements is simple: build defenses before threats emerge, not after. By using AI to systematically hunt for vulnerabilities across critical infrastructure, AWS and its partners aim to find and fix weaknesses before adversaries can discover them.

Key Takeaways for Cloud Security

• AI-Driven Efficiency: AI can reduce security log analysis time from hours to minutes, drastically increasing SecOps productivity.
• Scale as a Strength: Analyzing trillions of network flows allows for a collective defense mechanism where one detection protects all customers.
• Specialized Models: The shift toward specialized models like Claude Mythos Preview shows that general AI is evolving into high-performance tools for coding and cybersecurity.

The Future of Infrastructure Protection

As AI becomes more capable, the tools used to defend critical infrastructure must evolve in tandem. The integration of Project Glasswing and the deployment of advanced reasoning models mark a transition from reactive security to a predictive model. By prioritizing the discovery of vulnerabilities before they are weaponized, AWS is setting a new standard for how the industry secures the digital landscape.