OpenAI ‘ethically hacked’ with help of Anthropic’s Claude chatbot

US cybersecurity researchers who conducted hack say ‘scope of what we could theoretically access was huge’