Anthropic's AI security rollout

Anthropic initially released Project Glasswing to about 50 partners in April to test the model for cybersecurity flaws.

1. Frame 1Anthropic opens Project Glasswing to 150 partners as power, water, health, and government teams test AI for security flaws.
2. Frame 2The rollout expands beyond April's 50-partner test into infrastructure sectors where software bugs can interrupt essential services.
3. Frame 3New organizations must meet security requirements before gaining access, putting a gate between Mythos and sensitive codebases.
4. Frame 4The expanded pool spans more than 15 countries, including power, water, healthcare, communications, and hardware.
5. Frame 5Mythos has been described as identifying thousands of zero-day vulnerabilities across several weeks of code scanning.
6. Frame 6The pressure point is whether wider access fixes flaws faster while keeping powerful cyber tools under control.