OpenAI’s ChatGPT Rejects 250,000 Requests to Generate Deepfake Images of Candidates

OpenAI's ChatGPT denied 250,000 requests to generate deepfake images of candidates in the month leading up to the election. Instead, the AI fielded questions about voting and guided users to official sources. While ChatGPT took this approach, some other chatbots chose differently.

Artificial Intelligence and Election Season

As the presidential election drew near, OpenAI faced a surge in requests to produce fake images of candidates using its AI technology. OpenAI had previously implemented safeguards in its AI products to prevent misuse, including the creation of deepfakes or chatbots posing as political figures.

Guarding Against Misinformation

Concerns arose about AI potentially causing chaos during the election campaign by generating deepfakes and spreading conspiracy theories online. For instance, in January, voters in New Hampshire received deepfake robocalls impersonating President Joe Biden, discouraging them from participating in the state's presidential primary.

Rejecting Requests for Deepfake Images

OpenAI disclosed that ChatGPT turned down an estimated 250,000 requests to generate images of various candidates using DALL-E, the company's AI art generator, in the month leading up to the election. The AI system prevented the creation of fake images of President-elect Trump, Vice President Harris, Vice President-elect Vance, President Biden, and Governor Walz.

Guiding Users to Official Voting Information

OpenAI had ChatGPT focus on answering logistical queries about voting by directing users to CanIVote.org, a US voting information website managed by the National Association of Secretaries of State. Approximately 1 million responses directed users to the voting site in the month before November 5.

Providing Election Results Information

On Election Day, OpenAI ensured that ChatGPT responded to inquiries about election results by redirecting users to trusted news sources like the Associated Press and Reuters. The AI system included this message in around 2 million responses on Election Day and the subsequent day.

Unlike some chatbots, such as Elon Musk's Grok AI, which displayed enthusiasm for Trump's victory, ChatGPT refrained from expressing political opinions on candidates. OpenAI's efforts aimed to maintain neutrality and provide accurate information during the critical election period.

—————————————————————————————————————————————
By: kbalevic@businessinsider.com (Katie Balevic)
Title: OpenAI’s ChatGPT Rejects 250,000 Requests to Generate Deepfake Images of Candidates
Sourced From: www.businessinsider.com/chatgpt-openai-denied-requests-generate-images-candidates-election-trump-harris-2024-11
Published Date: Sat, 09 Nov 2024 20:28:54 +0000