OpenAI's ChatGPT has been at the center of attention leading up to the recent presidential election, as the artificial intelligence company revealed that it denied 250,000 requests for deepfake images of candidates in the month before Election Day. The company stated that its AI products have safeguards in place to prevent abuse, such as the creation of deepfakes or impersonation of candidates by chatbots.
Concerns about AI-generated deepfakes and conspiracy theories circulating online during the campaign have been escalating, with reports of fake robocalls targeting voters in New Hampshire earlier this year. In response to these potential threats, OpenAI's ChatGPT rejected a significant number of requests to generate images of prominent political figures using its AI art generator, DALL-E.
In addition to denying deepfake requests, ChatGPT also directed users to official outside sources for voting information, such as the website CanIVote.org, which is operated by the National Association of Secretaries of State. The AI-based chatbot provided around 1 million responses guiding users to check the voting site in the month leading up to the election.
On Election Day, OpenAI instructed ChatGPT to refer users to reputable news organizations like the Associated Press and Reuters for information on election results. The company reported that approximately 2 million responses included this message on Election Day and the following day.