Anthropic scales Project Glasswing to 150 partners across 15 countries to hunt critical software flaws

What happened

Anthropic has expanded Project Glasswing to include 150 partners across more than 15 countries. These partners use Anthropic’s Claude Mythos Preview AI model to scan critical infrastructure software for security vulnerabilities. The initiative has already uncovered over 10,000 serious flaws. At the same time, Anthropic is marketing Claude Security, a commercial product aimed at automatically fixing these vulnerabilities.

Why it matters

This move doubles down on security risks in essential digital infrastructure by deploying AI at scale to find bugs that human teams may miss. Anthropic is monetizing both discovery and resolution, creating a feedback loop that pressures organizations to adopt AI-driven fixes to keep critical systems safe. For companies running important software—and anyone reliant on that infrastructure—the scale and reach of this project mean more rapid exposure of hidden vulnerabilities and a new option to patch them automatically.

Anthropic’s model-driven approach accelerates vulnerability detection while squeezing older, labor-intensive methods. This tightens incentives to integrate AI tools directly into security workflows, lowering risk but also shifting power toward vendors who control these systems. It also puts pressure on organizations to respond faster to vulnerabilities or face higher chances of compromise.

What to watch next

Track how quickly the partner network grows beyond 15 countries and whether large enterprises outside the initial group start adopting Claude Security fixes. The balance between scanning for vulnerabilities and offering commercial patches creates a potential conflict of interest worth monitoring. It’s important to see if the product can handle complex, high-stakes environments without creating new risks or blind spots.

Also watch for industry reaction, including whether regulators begin scrutinizing AI’s role in automated vulnerability detection and remediation. Anthropic’s approach could reshape how infrastructure security evolves—either by tightening defenses or concentrating more control over critical fixes in AI providers’ hands.

AI Quick Briefs Editorial Desk