Security · Updated 7 Oct, 09:39 am IST

Anthropic expands program letting security teams test powerful AI models

Why it matters for readers: It explains why some AI models are tested with fewer safeguards and how that can help secure important software.

Anthropic merged two initiatives into a revamped Cyber Verification Program (CVP) with three tiers and varying verification and security controls.1 2
Project Glasswing partners verified at least 129,000 vulnerabilities between April and July, while Anthropic's open-source scanning found about 5,500 more between April and October.1 2
More than 33,000 of the reported vulnerabilities were rated critical or high severity.1 2
The program gives vetted security teams reduced safeguards on Anthropic models including Claude Mythos, Claude Opus and Sonnet to help find vulnerabilities.1 2
Anthropic said the reported figures are likely an undercount and expects the true impact to be at least five times higher.1 2

Get a brief like this every morning

Uzha reads hundreds of sources and gives you the stories that matter for your work, with every source linked. Free.

7 Oct, 12:34 am · ET Tech
Anthropic opens its most powerful AI models to more security teams
7 Oct, 09:39 am · The Hindu Technology
Anthropic opens its most powerful AI models to more security teams