OpenAI says the NYT hired a prompt wizard to make ChatGPT spit out its own articles
Ourtake
Whether it counts as hacking or evidence-gathering is now a question for the court, and the model's party trick is still on the record.
What happened
In a legal filing in its copyright dispute with The New York Times, OpenAI alleges the Times paid someone who used deceptive prompts, which violate OpenAI's terms, to make its GPT models generate copies of NYT articles. OpenAI says the regurgitation was a rare bug it has committed to fixing. The Times's lawyer says the activity was a search for evidence of copyrighted content.