Anthropic Expands Claude Access for Vetted Cyber Teams as Glasswing Finds 129,000 Flaws
Anthropic expands access to its advanced AI models for vetted cybersecurity professionals, allowing them to test with reduced safeguards and blocking classifiers. The company says Project Glasswing found at least 129,000 verified software vulnerabilities between April and July 2026.
Broader access to powerful models for security testing could help vetted defenders find weaknesses, while the reported vulnerability count underscores the scale of potential software risk.
