OpenAI Blocks Iranian Influence Operation Using ChatGPT for U.S. Election Content
OpenAI announced on Friday that it had taken action against a group of accounts associated with an Iranian covert influence operation. The operation utilized ChatGPT to create content that targeted various topics, with a focus on the upcoming U.S. presidential election.
Identifying the Covert Operation
OpenAI revealed that the content produced by ChatGPT included comments in both English and Spanish, which were posted on various social media accounts. Some of these comments were generated by rephrasing comments made by other users on social media platforms.
Content and Tactics
The articles created by the operation focused on U.S. politics, global events, and other relevant topics. They were published on multiple websites posing as progressive and conservative news outlets, indicating an attempt to reach audiences across the political spectrum.
Response and Detection
Microsoft also highlighted Storm-2035 as a threat activity cluster targeting U.S. voter groups with polarizing messages on various issues. The operation utilized fake news and commentary sites like EvenPolitics, Nio Thinker, Savannah Time, Teorator, and Westland Sun, which were observed plagiarizing content from U.S. publications.
Shifting Tactics
The propaganda network has reportedly adapted its tactics to avoid detection, using non-political posts and spoofing entertainment and non-political news outlets. The use of compromised accounts and link redirections has also been observed in an attempt to deceive users and evade detection.
Google's Threat Analysis Group has also detected and disrupted Iranian-backed spear-phishing campaigns targeting high-profile individuals in Israel and the U.S., including those associated with the U.S. presidential campaigns.
