OpenAI shuts down Iranian operation engaged in US election interference

Accounts created content that appeared to be from liberal and conservative-leaning users but did not receive high engagement

OpenAI in May said that networks from Russia, China, Iran and Israel had tried using the company’s AI products to enchance their propaganda efforts. Reuters

OpenAI removed a network of Iranian accounts using its ChatGPT chatbot to try to wage a foreign influence campaign on the US presidential election by generating long-form articles and social media comments on Friday.

The accounts created content that appeared to be from liberal and conservative-leaning users, including posts suggesting that former president Donald Trump was being censored on social media and was prepared to declare himself king of the US.

Another described Vice President Kamala Harris’s selection of Tim Walz for her running mate as a “calculated choice for unity”.

The influence campaign, which also included posts about the Israel-Gaza war, the Olympic Games in Paris, and fashion and beauty subjects, does not appear to have received significant audience engagement, said Ben Nimmo, principal investigator on OpenAI’s Intelligence and Investigations team, in a press briefing.

The Iranian operation marks the latest suspicious social media effort that used AI only to fail to get much traction, a possible indication that foreign operatives are still working on how to capitalise on AI tools that can quickly spit out convincing writing and images for little to no cost.

Microsoft this month alleged groups connected with Iran's government have increased cyber-influence efforts ahead of the US presidential election.

“They've laid the groundwork for influence campaigns on trending election-related topics and begun to activate these campaigns in an apparent effort to stir up controversy or sway voters – especially in swing states,” a blog post from Microsoft's general manager for threat analysis Clint Watts said.

Earlier this year, Meta said it had removed hundreds of Facebook accounts associated with influence operations from Iran, China and Russia, some of which relied on AI tools to spread disinformation.

Google said in a research report this week that it had observed a government-supported Iranian hacking group targeting “high-profile users in Israel and the US, including current and former government officials, political campaigns, diplomats, individuals who work at think tanks, as well as NGOs and academic institutions that contribute to foreign policy conversations”.

OpenAI said it had identified a dozen accounts on X and one on Instagram involved in the effort.

The disclosure comes Mr Trump's campaign accused Iran of hacking after Politico reported it had begun receiving emails in July containing internal documents from an anonymous account.

The FBI has opened an investigation into the alleged hack. Iran has denied involvement.

The US intelligence community has consistently warned about foreign governments trying to shape Americans’ opinions, with the Office of the Director of National Intelligence in July saying that Iran, Russia and China were recruiting people in the US to try spreading their propaganda.

OpenAI in May said that networks from Russia, China, Iran and Israel had tried using the company’s AI products to enhance their propaganda efforts.

It said the networks it disrupted had used AI to generate text and images in a larger volume than otherwise would have been possible by human creators, helping the content appear more authentic.

However, these campaigns also failed to generate significantly more engagement, according to the start-up.