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.