Exclusive: Anthropic’s Claude 3.7 Sonnet is the most secure AI model yet, an independent test finds

New independent research by the Holistic AI, a British firm that tests AI models, suggests Anthropic’s new Claude 3.7 Sonnet AI model cannot be persuaded to jump its built-in guardrails, making it the most secure AI model yet released.

Photo illustration by Cheng Xin—Getty Images