OpenAI has blocked more than 250,000 attempts to use its platform for creating deepfakes of US election candidates, the company announced. These requests, which aimed to generate AI-created images of prominent political figures like Donald Trump, Joe Biden, Kamala Harris, and their respective vice-presidential nominees, were denied as part of OpenAI’s proactive safety measures ahead of the election.
In a blog post published on Friday, OpenAI explained that these precautions were essential to prevent the misuse of its tools during the election cycle. The company stressed that these safeguards are part of a broader initiative to stop its AI models from being exploited for harmful or misleading purposes.
“These safeguards are critical in an election context, where the potential for deception is high,” OpenAI’s update stated. The company also reassured the public that there has been no indication that any influence operations related to the US election have successfully spread on its platforms.
Earlier this year, OpenAI disrupted a politically motivated influence campaign by a group called Storm-2035, which was attempting to use the platform to produce content about US politics and pose as both conservative and progressive media outlets. Following the disruption, related accounts were banned from OpenAI’s systems. In October, OpenAI revealed it had successfully shut down over 20 global influence operations attempting to misuse its tools for deceptive purposes.
However, OpenAI’s report clarified that none of the election-related operations targeted at its platforms were able to achieve viral reach.