OpenAI Says It Hosted ChatGPT-Powered ‘Iranian Influence Operation’

OpenAI said on Friday it had caught an “Iranian influence operation” using ChatGPT. The group, known as Storm-1679, produced articles and social media commentary to shape public opinion about Vice President Kamala Harris and former President Donald Trump, according to OpenAI. In addition to targeting the 2024 US presidential candidates, OpenAI said Storm-1679 also created content about Israel’s invasion of Gaza and its participation in the 2024 Olympics, the rights of US-based Latinx communities, Venezuelan politics, and and Scottish independence from the UK.

Most of the posts and articles seen by OpenAI received little footage from real people, the company said. However, it described the incident in detail on its blog, writing that it had found X accounts (formerly of Twitter) representing conservatives and progressives and using hashtags such as “#DumpTrump” and “#DumpKamala.” Storm-1679 has also tapped at least one Instagram account to distribute AI-generated content, per OpenAI.

OpenAI has previously described “state-coordinated threat actors” using its tools, but this is the first time it has uncovered a specific campaign to disrupt elections using ChatGPT.

OpenAI said it responded to the alleged discovery by banning “a collection” of accounts that created the content; the company also said it “shares threat intelligence with government, campaign and industry stakeholders.” The company did not name those involved directly, but shared screenshots of several posts. Those screenshots had view counts ranging from 8 to 207 and no likes.

OpenAI said Storm-1679 also shared articles generated by ChatGPT across several websites that “appeared to be consistent and consistent news sources.” The company added, “Most of the social media posts we identified received few or no likes, shares or comments. Similarly we did not find any indications that the web articles were shared on social media.”

An August 6 report from Microsoft described Storm-2035 similarly — as an Iranian network with “four websites masquerading as news outlets.” According to Microsoft, the network created “polarizing” posts about the election, LGBTQIA+ rights, and Israel’s attack on Gaza.

Reports of foreign internet interference in US elections are now commonplace. Microsoft’s August 6 report, for example, also detailed a phishing attack targeting an unnamed, “high-level” US campaign official. Shortly after Microsoft dropped the report, the Trump campaign announced that “outside sources” had stolen some of its emails and documents in an attempt to influence the 2024 presidential election. Eight years earlier, a Russian-linked hacking group known as Guccifer 2.0 had compromised the Democratic National Committee’s emails in a similar phishing attack; they eventually leaked thousands of DNC emails and documents before the 2016 Democratic National Convention.

Under a tidal wave of pressure from lawmakers, major technology companies have launched various initiatives over the years to respond to such incidents. Their efforts include meme fact-checking, wishful thinking, temporary bans on political ads, the “war room,” and collaboration with rivals and police alike.