OpenAI 'ethically hacked' with help of Anthropic's Claude chatbot

Here's an important detail that the title of the article obfuscates, "the researchers said they were largely using OpenAI’s own cutting edge GPT-5.6 Sol model to carry out the hack".