OpenAI blocks AI propaganda

OpenAI has found and banned a group of ChatGPT accounts linked to a renowned Iranian influence operation—called Storm-2035—that have been using ChatGPT to generate manipulative content designed to influence public opinion about a range of topics, including the US presidential elections.

It discovered the operation after reading the “Microsoft Threat Intelligence” report, which was released last week. The report highlighted that the group has been trying to influence US elections since 2020, by creating fake news and distributing it on social media.

Storm-2035 has been using ChatGPT to create posts, comments, and long-form blogs—including one that said X was censoring Donald Trump’s tweets—and has 12 X (formerly Twitter) accounts and 1 Instagram account which they’ve been using to spread their AI-generated propaganda messages.

Luckily, OpenAI didn’t find any evidence to suggest that this content was shared with many, establishing that the majority of its social media activity received minimal shares, likes, or comments, which is common with these cheap and fast campaigns.