White Hacking by AI: Anthropic Launches Project Glasswing for Cyber Defense

The application spectrum of Enterprise models is narrowing to high-margin niches. On April 7, Anthropic introduced Claude Mythos Preview—a closed research mode under Project Glasswing.

The project's goal is defensive cybersecurity. The model is specifically fine-tuned for vulnerability scanning, architectural auditing, and hardening critical software. In a world where hackers are already heavily using open models to generate exploits, businesses need automated AI auditing. The launch of Glasswing confirms the "Expert AI" trend: clients are willing to pay not for creative texts, but for rigorous, mathematically provable code analysis that prevents million-dollar losses from hacks.

Source: Anthropic / Reuters