Over 250,000 requests for US election deepfakes were rejected by ChatGPT, says OpenAI

The recent US election was the first time voters could refer to ChatGPT as a source for all things politics and voting, and now OpenAI just how many voters used the AI chatbot for this matter.

With the rise of AI and misinformation online, OpenAI took action before the election to prevent this kind of thing from spreading online by banning people from using ChatGPT to impersonate election candidates or governments.

It also prevented users from being able to distort information on voting or put people off from voting thanks to its partnership with the National Association of Secretaries of State to give the correct information.

The chatbot also referred people to the nonpartisan website CanIVote.org which had all the important details about voting - and one million ChatGPT responses were directed to this page, according to a blog post in the month before the election.

Meanwhile, AI-generated images made using the generator DALL-E were digitally watermarked to make them easier to spot.
The blog post also stated how over a staggering 250,000 requests for deepfakes of President-elect Donald Trump, Vice President Kamala Harris, Vice President-elect JD Vance, and Governor Tim Walz were rejected by ChatGPT.

Photo by LIONEL BONAVENTURE/AFP via Getty Images

Both on and the day after the election, two million ChatGPT responses recommended Associated Press and Reuters for users to find out the election results.

Although these measures were in place, there are still OpenAI misinformation concerns from the Bipartisan Policy Center after it put election-related questions to ChatGPT.

It noted the importance "to exercise discretion regarding its applications, especially when there are significant implications on our democracy," and believes further advancements can be made before the 2026 midterm elections.

"It is easy to mislead users when solely relying on unconfirmed sources, such as ChatGPT and other chatbots, for answers," according to Bipartisan Policy Center's report.

"The bot has limitations of prompt length and information training and often does not answer with complete or consistent information. We would caution users to check ChatGPT answers with reliable resources such as government websites or their local election boards."

How to join the indy100's free WhatsApp channel

Sign up to our free indy100 weekly newsletter

Have your say in our news democracy. Click the upvote icon at the top of the page to help raise this article through the indy100 rankings