These concerns come after OpenAI announced in January that it would ban people from using its technology to create chatbots that imitate political candidates or provide false information related to voting. This is one of the reasons why we announced this. The company also said it does not allow applications to be built for political campaigning or lobbying purposes.
Although the Kennedy chatbot page does not expose the underlying model that powers the chatbot, the site's source code connects the bot to LiveChatAI. LiveChatAI is a company touting its ability to provide businesses with customer support chatbots powered by GPT-4 and GPT-3.5. His website for LiveChatAI describes the bot as “leveraging the power of ChatGPT.”
LiveChatAI co-founder Emre Elbeyoglu, in response to a question about which large language models are used in the Kennedy campaign bot, said in an emailed statement Thursday that the platform does not support GPT-3.5 or GPT-4. In addition, “we use a variety of technologies such as Llama and Mistral.” “Due to our commitment to client confidentiality, we cannot confirm or deny the details of our client's use,” Elbeyoglu said.
OpenAI spokesperson Nico Felix told WIRED on Thursday that the company has “no indication” that the Kennedy campaign's chatbot is directly building on its services, but that LiveChatAI will He suggested it might be using one of the company's models. Since 2019, Microsoft has reportedly invested over $13 billion in OpenAI. OpenAI's ChatGPT model was later integrated into Microsoft's Bing search engine and the company's Office 365 Copilot.
On Friday, a Microsoft spokesperson confirmed that the Kennedy chatbot “leverages the power of the Microsoft Azure OpenAI Service.” Microsoft said its customers are not bound by OpenAI's terms of service and the Kennedy chatbot does not violate Microsoft's policies.
“In limited testing of this chatbot, we have demonstrated its ability to generate answers that reflect the intended context while taking appropriate care to prevent misinformation,” the spokesperson said. “If we find an issue, we work with our customers to understand and guide them towards usage consistent with those principles. In some scenarios, this could result in the termination of a customer's access to our technology. It may be possible to connect.”
OpenAI did not respond to WIRED's request for comment on whether the bot violated its rules. Earlier this year, the company blocked the developer of Dean.bot, a chatbot built on OpenAI's model that mimics Democratic presidential candidate Dean Phillips and answers voter questions.
As of late Sunday afternoon, the chatbot service was no longer available. The page is still accessible on Kennedy's campaign site, but the embedded chatbot window displays a red exclamation mark icon and simply says “Chatbot not found.” WIRED reached out to Microsoft, OpenAI, LiveChatAI, and the Kennedy campaign for comment on the chatbot's apparent removal, but did not immediately receive a response.
Chatbots are prone to hallucinations and hiccups, making their use in political contexts controversial. Currently, OpenAI is the only major large-scale language model that explicitly prohibits its use in campaigns. Meta, Microsoft, Google, and Mistral all have terms of service, but they don't directly address politics. And given that campaigns can access GPT-3.5 and GPT-4 through third parties with no apparent repercussions, there are few restrictions.
“I can tell you that OpenAI does not allow its tools to be used in elections or, on the other hand, allow its tools to be used in election campaigns,” Woolley said. “But on the other hand, we now have access to these tools fairly freely. Given the decentralized nature of this technology, one wonders how Open AI will actually enforce its own policies. is needed.”