Anthropic says it found 5,500 verified vulnerabilities in April-October and Glasswing partners found 129K+ in April-July; 33K+ were critical or high severity (Reuters) — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic found 5,500 verified vulnerabilities in its AI models between April and October, while Glasswing partners uncovered 129,000+ vulnerabilities from April through July, with 33,000+ classified as critical or high severity.
- Anthropic is expanding a program that allows vetted cybersecurity professionals to test its most powerful AI models with fewer safety restrictions.
Why it matters: Anthropic is simultaneously publishing vulnerability counts and loosening access for external testers, meaning the 33,000+ critical and high-severity findings are likely to grow — putting pressure on enterprise customers and regulators evaluating whether frontier models can be safely deployed.
Ask SkimNews
