ChatGPT Blocks 250,000 Requests for Deepfake Political Images Pre-Election

As the presidential election approached this week, OpenAI received a deluge of requests to produce counterfeit images of political figures.

Earlier this year, the company emphasized that its AI tools are equipped with 'safeguards' to prevent misuse, including the creation of deepfake images or impersonation of political candidates via chatbots.

This announcement emerged amidst concerns about the potential chaos AI could cause during the election campaign, particularly regarding the spread of deepfakes and conspiracy theories online. In January, fake robocalls featuring President Joe Biden's voice, intended to dissuade New Hampshire residents from participating in the state’s primary, highlighted these risks.

According to OpenAI, ChatGPT turned down approximately 250,000 requests to craft images of political figures using DALL-E, their AI-powered art generator, during the month leading up to the election.

"In the weeks before Election Day, it's estimated that ChatGPT declined over a quarter of a million requests aimed at generating DALL-E visuals of individuals such as President-elect Trump, Vice President Harris, Vice President-elect Vance, President Biden, and Governor Walz," declared OpenAI in a blog update on Friday.

AI’s Role in Election-Related Information and Neutrality

OpenAI had previously committed to guiding ChatGPT to address voting-related inquiries by directing users to CanIVote.org, a comprehensive voting information platform managed by the National Association of Secretaries of State. Over the month leading up to November 5, ChatGPT reportedly delivered nearly one million responses that encouraged users to verify details on this voting site.

Furthermore, on Election Day, ChatGPT was programmed to answer queries about election outcomes by referring users to reputable media sources like the Associated Press and Reuters.

OpenAI reported that nearly two million replies from ChatGPT on the day of and the day following the election included directions to these news outlets, underscoring their commitment to providing reliable election information.

Maintaining Objectivity in Political Discourse

In contrast to other AI systems, such as Elon Musk's Grok AI, which demonstrated a preference by celebrating Trump's victory, ChatGPT refrained from expressing political opinions about candidates.