Iranian Influence Operation Manipulates U.S. Election Discourse

EXECUTIVE SUMMARY

OpenAI uncovered and acted against a covert Iranian influence operation known as Storm-2035 that was using ChatGPT to generate politically charged content related to global events, the U.S. presidential election, and other sensitive topics. This operation was part of a broader trend of foreign interference targeting political processes in the U.S., with the goal of manipulating public opinion through fabricated content and misinformation. The campaign appeared to focus on polarizing issues such as the conflict in Gaza, Israel's involvement in the Olympics, U.S. politics, and the rights of Latinx communities, both in English and Spanish. While the operation attempted to engage audiences through various social media platforms and websites posing as both conservative and progressive news outlets, its efforts failed to achieve significant engagement, as most posts received little to no interactions.[emaillocker id="1283"]

The Storm-2035 operation leveraged ChatGPT to generate long-form articles and short comments that were posted on X (formerly Twitter) and Instagram. The content covered diverse topics but primarily focused on politically sensitive matters, including the U.S. presidential election and the Israel-Hamas conflict. The operation aimed to create a divide among American voters by generating commentary on both sides of the political spectrum and sharing content through progressive and conservative-styled news outlets. In addition to the political posts, Storm-2035 occasionally mixed in fashion and beauty-related content, likely as an attempt to make the accounts appear more authentic and gain a broader following. OpenAI’s models were used to rewrite comments previously posted by other users to evade detection. The campaign, which had a minimal impact on audience engagement, was assessed using Brookings' Breakout Scale and categorized as Category 2—a level that indicates the operation did not achieve widespread reach. The operation’s use of ChatGPT for covert influence tactics echoes previous cases reported in May 2024, and similar methodologies have been noted in foreign influence operations conducted by Russian and Iranian threat actors.

Although Storm-2035 failed in significantly influencing its target audience, the incident highlights the growing use of AI tools in covert influence operations, especially in politically sensitive contexts like elections. Despite its lack of widespread engagement, OpenAI's prompt detection and removal of the accounts underscore its commitment to preventing abuse of its AI models for malicious purposes. As these kinds of operations become increasingly sophisticated, they pose a serious risk to democratic processes by amplifying misinformation, deepening political divides, and eroding public trust. Moving forward, OpenAI continues to enhance its efforts to monitor and address such activity, collaborating with government, industry, and civil society to ensure that its AI technology is not misused. The case of Storm-2035 serves as a reminder of the critical need for vigilance in detecting and mitigating foreign influence campaigns, especially as they adopt new technologies like generative AI to amplify their efforts.

THREAT PROFILE:

REFERENCES:

The following reports contain further technical details:

https://thehackernews.com/2024/08/openai-blocks-iranian-influence.html