OpenAI, the creator of artificial intelligence (AI) chatbot sensation ChatGPT, has suspended the account of a third-party developer for violating its policies by using AI technology to make a conversational bot in support of the presidential campaign of Rep. Dean Phillips (D-Minn.).
The developer account belonged to Silicon Valley startup Delphi, reports The Washington Post. It had been contracted by We Deserve Better, a Super PAC founded by tech entrepreneurs to back the longshot Phillips campaign, to create “Dean.Bot” – a ChatGPT-powered website that could chat with voters in real time and answer their questions about the candidate’s positions.
While Dean.Bot included disclaimers that it was an AI tool, OpenAI determined the application still violated its ban on using its conversational models for political campaigning or impersonating a person without consent. On Friday night (Jan.19) , after The Washington Post published an article detailing the Dean Phillips chatbot, OpenAI suspended Delphi’s developer account and access to ChatGPT. In response, Delphi took Dean.Bot offline.
OpenAI takes action
An OpenAI spokesperson explained the account suspension by stating: “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 move marks the first known enforcement action taken by OpenAI over perceived misuse of ChatGPT and its underlying AI capabilities in the context of a US election. While Delphi and We Deserve Better viewed Dean.Bot as an innocuous, transparent tool to engage and educate voters, AI ethics experts have raised alarms about the potential abuse of such technologies to spread disinformation and manipulate the democratic process.
By targeting Delphi over Dean.Bot, OpenAI is drawing an early line in the sand on AI chatbots in campaigns.
Earlier this month, OpenAI outlined its strategy for safeguarding its powerful large language and image models from being weaponized by bad actors during the many elections set to be held in 2024.
Featured Image: Sanket Mishra/Pexels
