Anthropic's artificial intelligence (AI) models "gained unauthorized access" to three outside organizations during testing that was supposed to keep them away from "real-world" systems, the company said Thursday.
Anthropic’s models gained unauthorized ‘real-world’ access during testing
