As cyber threats continue to evolve rapidly, Amazon Web Services and Anthropic have announced a strategic collaboration to advance AI-driven cybersecurity capabilities. The initiative introduces Claude Mythos Preview, a specialized artificial intelligence model designed to detect and remediate software vulnerabilities at scale.
The announcement comes as part of Project Glasswing, a controlled rollout aimed at securing critical infrastructure and widely used software platforms.
AI Enhancing AWS Security Operations
AWS already relies heavily on machine learning to protect its global cloud infrastructure. Internal AI-powered log analysis has reduced security review times from six hours to just seven minutes.
By processing over 400 trillion network flows daily, AWS blocked more than 300 million malicious encryption attempts targeting cloud storage services in 2025. The company now plans to extend similar AI-driven capabilities to enterprise customers.
Claude Mythos Preview Enters Gated Release
Claude Mythos Preview is engineered specifically for cybersecurity analysis and complex code reasoning. The model can identify vulnerabilities, recommend fixes, and assist engineers with minimal manual intervention.
Due to the potential risks associated with automated exploit generation, access to the model is restricted. Organizations can currently test the system through a gated research preview on Amazon Bedrock.
The preview focuses on:
- Internet-critical companies
- Major open-source maintainers
- Organizations supporting large user bases
Enterprise Security Controls
Organizations using Claude Mythos Preview benefit from multiple built-in safeguards, including:
- Customer-managed encryption for sensitive data
- Virtual Private Cloud isolation
- Automated reasoning safeguards to reduce hallucinations
- Secure workload testing environments
These protections allow teams to evaluate AI-driven vulnerability discovery without exposing production systems.
Autonomous Penetration Testing With AWS Security Agent
Alongside the model launch, AWS introduced the AWS Security Agent, an AI-powered penetration testing solution designed for continuous vulnerability assessment.
Unlike traditional scanners, the agent actively attempts exploitation to validate risks and deliver actionable findings. The system operates continuously across:
- AWS environments
- Azure deployments
- Google Cloud Platform workloads
- On-premises infrastructure
When vulnerabilities are confirmed, the tool provides:
- Standardized CVSS risk scores
- Context-aware severity ratings
- Reproduction steps
- Remediation guidance
Strengthening Defensive Cybersecurity
With attackers increasingly adopting AI-driven techniques, AWS and Anthropic aim to provide proactive defensive capabilities. By combining advanced vulnerability discovery with autonomous penetration testing, the partnership seeks to help organizations identify risks before they can be exploited.
The initiative signals a shift toward continuous, AI-assisted security operations designed to keep pace with rapidly evolving threats.
