OpenAI bans a third-party developer for making a US presidential candidate GPT-based chatbot

Just a few days ago, the generative AI company OpenAI revealed how it will regulate the use of its ChatGPT and DALL-E AI models during the current 2024 elections. One of those new policies was that OpenAI would not allow its tools and services to be used to create simulated versions of real election candidates.

This weekend, OpenAI enforced that particular policy by banning a developer who had used ChatGPT to create a chatbot that simulated US House of Representatives member Dean Phillips of the third district in Minnesota. Phillips is a declared candidate for US President in the Democratic Party, in a long-shot campaign against current President Joe Biden.

The Washington Post reports that "Dean.Bot" was created by the developer company Delphi. It had received funding from We Deserve Better, a Super PAC made to help support Phillips" campaign. While the bot had disclaimers in place stating this was an AI tool, it still violated OpenAI"s rules against making such chatbots.

On Thursday, the We Deserve Better super PAC asked Delphi to no longer use ChatGPT for "Dean.Bot" and instead use open source-based AI models for the chatbot. However, on Friday, Delphi shut down the chatbot entirely after ChatGPT suspended the developer. The story includes a quote from a ChatGPT spokesperson:

“Anyone who builds with our tools must follow our usage policies. We recently removed a developer account that was knowingly violating our API usage policies which disallow political campaigning, or impersonating an individual without consent.”

The story also mentions that one of the super PAC"s co-founders, Matt Krisiloff, was previously the chief of staff to OpenAI"s co-founder and current CEO Sam Altman. Krisiloff claims that Altman has no involvement in the super PAC, but did admit he has met Rep. Phillips in the past.

Report a problem with article
Next Article

Get this huge 98-inch TCL TV for $2,000 off for a new all time low price right now

Previous Article

GenAI could make software much less buggy by 2028 with better unit tests