OpenAI says Iranian group using ChatGPT tried to sow division among U.S. voters
OpenAI recently revealed that it had taken action against a cluster of ChatGPT accounts that were found to be associated with an Iranian group aiming to create discord among U.S. citizens. The decision to ban these accounts came after the discovery of online content, including articles and social media posts, generated with the assistance of ChatGPT. Although the reach of this content was limited, it focused on contentious topics such as the U.S. presidential election, Israel's conflict in Gaza, and Israel's participation in the Olympic Games. The AI-generated materials were tailored to appeal to both progressive and conservative audiences.
Investigation Findings
OpenAI stated that the orchestrated operation did not succeed in engaging a substantial audience, with most social media posts receiving minimal likes, shares, or comments. The investigation further unveiled that AI-generated long-form articles were disseminated on platforms masquerading as legitimate news sources. Additionally, the banned accounts were responsible for crafting social media comments, in both English and Spanish, by instructing ChatGPT to rephrase existing comments posted by other users. Notably, the political content was intertwined with discussions on fashion and beauty, potentially to enhance credibility or attract a larger following, according to OpenAI.
Investigation Findings
OpenAI stated that the orchestrated operation did not succeed in engaging a substantial audience, with most social media posts receiving minimal likes, shares, or comments. The investigation further unveiled that AI-generated long-form articles were disseminated on platforms masquerading as legitimate news sources. Additionally, the banned accounts were responsible for crafting social media comments, in both English and Spanish, by instructing ChatGPT to rephrase existing comments posted by other users. Notably, the political content was intertwined with discussions on fashion and beauty, potentially to enhance credibility or attract a larger following, according to OpenAI.
Collaborative Efforts
OpenAI's internal probe coincided with the release of a report by Microsoft's Threat Analysis Center, which highlighted the utilization of various online strategies by groups associated with the Iranian government to interfere in the U.S. presidential election. Subsequently, Donald Trump's presidential campaign reported a security breach, attributing it to Iranian actors. Although the campaign did not furnish specific evidence, the incident underscored the ongoing cybersecurity challenges faced by political entities during electoral campaigns.
Covert Operation Unveiled
OpenAI linked the ChatGPT accounts to a covert initiative known as Storm-2035. A parallel investigation by Microsoft exposed the group's operation through four fabricated websites posing as American news platforms. These deceptive sites amplified divisive narratives on issues concerning LGBTQ rights and the Israel-Hamas conflict, with the primary objective of triggering chaos and exacerbating societal divisions among American voters in the lead-up to Election Day.
