OpenAI shuts down Iranian group’s ChatGPT accounts over U.S. election influence campaign

OpenAI has taken decisive action by banning a cluster of ChatGPT accounts connected to an Iranian influence operation. These accounts were posting material produced by AI concerning the U.S. presidential election, as the company explained in a blog post. The operation, which contained articles and social media posts, seemed to have had a small scope. 

This is not the first time that state-affiliated actors have tried to misuse the OpenAI’s ChatGPT. In May, the company prevented five different attempts to influence the public with fake media articles posted by AI bots. These efforts are part of a wider pattern in which state actors use new technologies to try to shape political processes. 

OpenAI responds swiftly to counter emerging threats

Earlier on, such groups used Facebook and Twitter as their major social media platforms. However, generative AI is now used to overwhelm social media platforms with fake information.  In response to these threats, OpenAI has been quick, similar to the “whack-a-mole” approach we have seen in social media platforms. As the accounts connected with such activities come to light, they are immediately blocked. This is important as the 2024 United States presidential election is slowly coming into the picture. 

OpenAI’s analysis of the recent batch of accounts is powered by a report from Microsoft Threat Intelligence that came out last week. The report named the group sponsoring these operations as the Storm-2035 group, which has been involved in interfering with the U.S. elections since 2020. 

Storm-2035 is reported to be an Iranian network with numerous domains that are meant to resemble genuine news websites. These sites target the American voter and present polarizing content pertaining to presidential candidates, gay rights, and the Israel-Hamas conflict. 

ChatGPT generates polarizing content on various political topics

As in other similar operations, the purpose of Storm-2035 is not to back an individual policy but to cause controversy among the population. According to OpenAI, the group was associated with five websites that posed as both liberal and conservative news sites. 

These sites had good domain names, such as “evenpolitics. com,” and used ChatGPT to write articles of a certain length. One of these articles stated that the social media network owned by Elon Musk, X, was blocking former U.S. President, Donald Trump’s tweets. 

OpenAI has identified a dozen of the X accounts and one Instagram account linked to the Storm-2035. These accounts employed ChatGPT to generate new political statements and then published these on these social media channels. For instance, a tweet falsely implicated climate change in ‘increased immigration costs’ while tagging the statement with #DumpKamala in reference to Vice President Kamala Harris.