OpenAI has banned the developer of Dean.Bot, an artificial intelligence (AI) powered bot that mimics US presidential candidate Dean Phillips. The Microsoft-backed AI startup said the developer violated company policy.
To those unaware, Phillips, 55, announced his bid for the presidency in October, arguing a younger generation should be given the opportunity to lead. While campaigning in New Hampshire during the weekend, Phillips described Biden as "unelectable and weak".
A report by The Washington Post points out this is the first action OpenAI has taken in response to what it believes is a misuse of its AI tools in a political campaign.
"Anyone who builds with our tools must follow our usage policies," OpenAI spokeswoman Lindsey Held said in a statement to Reuters.
"We recently removed a developer account that was knowingly violating our API usage policies which disallow political campaigning, or impersonating an individual without consent," Held added.
The ChatGPT-powered bot that impersonated Phillips, who is challenging US President Joe Biden for the Democratic party candidacy, is available on the dean.bot site.
https://t.co/ec2A4iekZZ worked on the basis of #ChatGPT - he answered voters’ questions and campaigned to vote for the representative of the US Democratic Party Dean Phillips.
— TheRevolutionReport (@TheRevolutionR1) January 22, 2024
The bot was backed by Silicon Valley entrepreneurs Matt Krisiloff and Jed Somers. The duo started a super PAC called We Deserve Better to support Phillips ahead of the impending New Hampshire primary.
The PAC had reportedly received $1 million (£0.79 million) from hedge fund manager Bill Ackman, who was willing to invest in Elon Musk-led X through his SPARC (special purpose acquisition rights company) structure last year.
The bot, which has been made available to outside developers, was built by AI start-up Delphi. OpenAI banned Delphi's account last week in response to a Washington Post story that linked OpenAI CEO Sam Altman and Bill Ackman to a new super PAC that supports Phillips.
A new super PAC backed by Silicon Valley insiders is mobilizing to spread Dean Phillips’s ideas with https://t.co/SGIrq9CCZj, an artificial intelligence tool that can chat like a real person — one of the first known uses of AI in a political campaign. https://t.co/GzwJ6STPvZ
— The Washington Post (@washingtonpost) January 19, 2024
The American AI company noted that it does not allow or support the use of its technology in political campaigns. Following the account suspension, Delphi removed Dean.Bot.
How technology can influence political campaigns
Delphi co-founder Dara Ladjevardian told The Post that the company "incorrectly" believed that OpenAI's terms of service would allow "a political action committee that supports Dean Phillips to create a clone of him using our platform".
According to Ladjevardian, the start-up "did not understand that ... [the super PAC] may not and did not coordinate with or seek permission from candidates they are supporting".
Moreover, Ladjevardian noted that he had refunded the super PAC and updated the company's terms of service to restrict engagement with political campaigns.
Dean.Bot's ability to converse with voters in real-time is an early use of AI that researchers have warned could harm upcoming elections.
The bot had a disclaimer that confirmed that it was an AI tool and not the real Dean Phillips. While the bot required that voters consent to its use, researchers told The Post that such technologies could trick people into accepting a dangerous tool, despite the presence of disclaimers.
We Deserve Better and other proponents, on the other hand, claim that bots can play a key role in educating voters if used appropriately. These bots can help voters learn more about a candidate in an entertaining way.
Without disclaimers, experts suggest the technologies could allow mass robocalls to voters who think they are talking to actual candidates. On top of that, they believe AI systems can help cybercriminals spread disinformation through fake websites.
The use of AI in the US elections is inevitable.
— Lorenzo Green 〰️ (@mrgreen) March 20, 2023
The question is, how far will they go to get your vote?
- Deepfakes
- Hyper-personalized ads & content.
- Data mining.
- Psychological profiling.
- AI-generated fake news.
They already have access to personal data...
Thoughts? pic.twitter.com/7TOowFOBaF
Deepfakes and AI-generated disinformation could disrupt elections around the world this year, with the US, EU, UK and India all gearing up to hold polls.
In line with this, AI expert Oren Etzioni last month said he expects a "tsunami of misinformation" ahead of the US presidential election.
Following OpenAI's prohibitions, Krisiloff told The Post that Delphi has removed ChatGPT from the bot and now relies on open-source technologies that support conversational capabilities.
