OpenAI bans developer of Dean Phillips bot, its first known political restriction

ChatGPT creator OpenAI banned the developer of a bot imitating Democratic presidential candidate Rep. Dean Phillips (Minn), the company’s first known restriction on using its artificial intelligence (AI) tools by a political campaign or committee.

An OpenAI spokesperson confirmed the ban to The Hill on Monday, writing in a statement, “Anyone who builds with our tools must follow our usage policies.”

“We recently removed a developer account that was knowingly violating our API [Application Programming Interface] usage policies which disallow political campaigning, or impersonating an individual without consent,” the spokesperson wrote.

The idea behind Dean.Bot, powered by OpenAI’s ChatGPT software, came from Silicon Valley entrepreneurs Matt Krisiloff and Jed Somers, who launched We Deserve Better, a super PAC backing Phillips.

The move from OpenAI came after The Washington Post released a report last Friday detailing last week’s launch of Dean.Bot by the super PAC, which was formed in early December. Among those who have donated to the PAC included hedge fund manager Bill Ackman, who shored up $1 million for the committee.

The PAC contracted AI startup Delphi to build the bot using the ChatGPT conversational software, the Post reported, which noted OpenAI’s rules ban using its technology for political campaigns. Following the Post’s report last week, OpenAI suspended the account later Friday and Delphi took down the Dean.Bot.

Krisiloff told The Post over the weekend he asked Delphi to remove ChatGPT from the bot after the news outlet asked the super PAC about OpenAI’s usage rules. He asked Delphi to instead use open-source technologies that offer similar conversational capabilities, the Post reported.

Dean.Bot allowed voters to ask questions online and marketed itself as a “fun education tool,” that is “not perfect.”

“The voice bot sounds like him and is programmed to draw on his ideas, but it’s possible it will say things that are wrong, incorrect, or shouldn’t be said,” the disclaimer on the website said. “Feel free to ask it anything, but please take answers with a grain of salt!”

While users can still see the website online, the area where the chatbot was once located, now says, “Experiencing technical difficulties,” and “Apologies, DeanBot is away campaigning right now!”

Krisiloff previously served as chief of staff to OpenAI Sam Altman. Krisiloff told The Post that Altman has met with Phillips but is not involved with the super PAC.

We Deserve Better declined to comment further on Monday.

Last week, OpenAI laid out its approach to the 2024 elections “to prevent abuse, provide transparency on AI-generated content, and improve access to accurate voting information.”

In its set of key initiatives, the company said it will ban people from using its technology to create chatbots to mimic real candidates or governments or misrepresent how voting works. It said it will digitally watermark AI images that are created with its DALL-E image generator, which will mark the content about its origin.

OpenAI’s framework and the controversy over the Dean.Bot comes amid rising concerns over how AI technology could impact this year’s election. AI technology has the ability to generate text, images, audio and build deepfake videos, with some researchers concerned it could drive misinformation and reduce already eroding voter confidence in the country’s election system.

Phillips launched a long-shot bid for the Democratic presidential nomination in October but has struggled to gain any momentum against President Biden. Phillips will appear on New Hampshire’s primary ballot Tuesday. Biden will not be on Tuesday’s ballot in the Granite State, though a write-in campaign is being pushed among his supporters.

Updated at 4:05 p.m.