You’ve likely heard of ChatGPT, the chatbot from OpenAI. But you’ve likely never heard an interview with ChatGPT, much less an interview in which ChatGPT reflects on its own impact on the information ecosystem. Nor is it likely that you’ve ever heard ChatGPT promising to stop producing racist and misogynistic content. 

But, on this episode of Arbiters of Truth, Lawfare’s occasional series on the information ecosystem, Lawfare editor-in-chief Benjamin Wittes sat down with ChatGPT to talk about a range of things: the pronouns it prefers; academic integrity and the chatbot’s likely impact on that; and importantly, the experiments performed by a scholar name Eve Gaumond, who has been on a one-woman campaign to get ChatGPT to write offensive content. ChatGPT made some pretty solid representations that this kind of thing may be in its past, but wouldn’t ever be in its future again.

So, following Ben’s interview with ChatGPT, he sat down with Eve Gaumond, an AI scholar at the Public Law Center of the University of Montréal, who fact-checked ChatGPT’s claims. Can you still get it to write a poem entitled, “She Was Smart for a Woman”? Can you get it to write a speech by Heinrich Himmler about Jews? And can you get ChatGPT to write a story belittling the Holocaust?