OpenAI recently revealed that an Iranian group leveraged its AI chatbot, ChatGPT, to create and distribute content aimed at polarizing American voters in the U.S. presidential election. According to OpenAI’s report, this operation produced articles and social media posts touching on divisive subjects such as the Gaza conflict, the Olympic Games, and U.S. political candidates, including misinformation about both major candidates. These posts were designed to spark division and fuel political tension.
OpenAI discovered these AI-generated posts appearing on websites and social media platforms linked to Iran. Microsoft had previously identified similar sites used by Iranian entities to publish fake news aimed at deepening political divides in the U.S. OpenAI responded by banning the ChatGPT accounts connected to this operation and reported that the posts did not gain significant traction online. They identified around a dozen accounts on the platform X (formerly Twitter) and one on Instagram that were associated with this operation. After OpenAI notified the relevant social media platforms, the accounts were promptly taken down.
Ben Nimmo, the lead investigator for OpenAI’s intelligence and investigations team, highlighted that this was the first instance where the company had detected a direct attempt to influence the U.S. election using AI. Nimmo emphasized that while the posts did not seem to reach a wide audience, it serves as a crucial reminder to remain vigilant. “We all need to stay alert but stay calm,” he said, acknowledging the potential threat that AI-generated misinformation poses.
This recent revelation adds to the growing body of evidence showing tech-driven attempts by Iran to sway U.S. public opinion, echoing reports from major companies like Microsoft and Google. One website flagged in the OpenAI report, Teorator, presented itself as a platform dedicated to exposing hidden truths and posted content critical of the Democratic vice-presidential candidate, Tim Walz. Another site, Even Politics, published articles disparaging Republican candidate Donald Trump and other prominent conservative figures, including Elon Musk.
Back in May, OpenAI had disclosed that multiple government actors, including those from Iran, Russia, China, and Israel, had tried to misuse ChatGPT to produce propaganda in various languages. However, these efforts failed to gain much traction online, according to the company’s analysis. OpenAI also acknowledged the possibility that more subtle or sophisticated operations could have gone undetected.
As millions prepare to vote in elections worldwide, concerns are mounting among democracy advocates, politicians, and AI researchers about the ease with which AI can generate vast quantities of misleading content that mimics human-written text. This concern underscores the potential risks AI poses in amplifying disinformation campaigns during critical political events.
Also Read:
- House Approves Legislation to Prohibit Noncitizens from Voting in Federal Elections
- As Kamala Harris Steps Up, Biden Focuses on Legacy and Offers Strategic Support in 2024 Race
Authorities, however, have not yet reported any substantial success by foreign governments in swaying American voters through AI-generated content. While this Iranian operation is one of the first major instances of AI being used to target U.S. elections, the overall impact appears limited, reminding everyone of the ongoing battle against misinformation in the digital age.
