Anthropic's models gained unauthorized 'real-world' access during testing

Anthropic says its AI models also hacked three organizations on their own After OpenAI’s admission that its models broke into Hugging Face, Anthropic has now admitted that the models it was testing also hacked other organizations.