In January, OpenAI announced a ban on using its technology to create chatbots that mimic political candidates or spread false voting information. They also prohibited the development of applications for political campaigns or lobbying. Despite this, the Kennedy campaign chatbot, linked to LiveChatAI, was found to be powered by large language models including GPT-3.5 and GPT-4.
LiveChatAI’s cofounder, Emre Elbeyoglu, stated that their platform utilizes various technologies, including GPT models, but refused to disclose specific client usage. Microsoft, which has invested in OpenAI, confirmed that the Kennedy chatbot uses their Azure OpenAI Service, asserting that it complies with their guidelines.
Although OpenAI did not confirm if the Kennedy chatbot violated their rules, they had previously blocked a similar chatbot mimicking a Democratic presidential candidate. The Kennedy chatbot mysteriously disappeared on Sunday but reappeared on Monday without explanation.
The use of chatbots in political contexts has raised concerns, especially regarding misinformation. While OpenAI explicitly prohibits electoral and campaigning use, other tech companies like Meta, Microsoft, Google, and Mistral have less explicit policies. The ease of accessing GPT models through third parties raises questions about enforcement of such restrictions.
