Anthropic Expands Claude Access for Vetted Cyber Teams as Glasswing Finds 129,000 Flaws

Anthropic expanded its Cyber Verification Program to let vetted cybersecurity teams test advanced AI models with fewer safeguards after Project Glasswing and related scanning efforts reported discovering large numbers of software vulnerabilities (projected to be significantly higher than reported), including tens of thousands rated critical/high; the write-up also notes limited observed exploitation so far (2 of 300 sampled vulnerabilities) and warns that AI-generated fixes can introduce additional security risks.