OpenAI recently uncovered and disrupted five influence operations that utilized artificial intelligence tools to spread political messaging. These operations originated from China, Iran, Israel, and two from Russia. The primary use of AI in these operations was to generate text, such as social media posts, comments, and productivity tasks like debugging code.
Despite their attempts, none of the influence operations were highly effective. On the Brookings Breakout Scale, which measures the impact of influence operations on a scale of 1 to 6, none of them scored higher than a 2. A score of 1 indicates that the campaign spread within a single community or platform, while a score of 6 would involve triggering concrete actions like violence. A score of 2 means the operation spread across multiple communities on one platform, or one community across multiple platforms.
The influence operations showcased a diversity of tactics but had striking similarities in their methodologies. For example, Spamouflage, originating from China, used OpenAI tools to debug its code, research social media activity, and post content on various platforms in multiple languages like X, Medium, and Blogspot.
Another operation, Bad Grammar, identified as a threat from Russia, primarily operated on Telegram and targeted individuals in Eastern Europe and the United States. It used AI to debug code for running a Telegram bot and posting political comments in both Russian and English.
Doppelganger, a second Russian group, utilized AI to post comments on platforms like X and 9GAG in five European languages, generate headlines, as well as translate, edit, and convert news articles into Facebook posts.
The International Union of Virtual Media (IUVM), an Iranian entity, employed AI for generating and translating articles, headlines, and website tags on its platform. Lastly, Zero Zeno, an operation run by Stoic, a Tel Aviv-based company, used OpenAI to create articles and comments for various websites like Instagram, Facebook, and X.
Stoic’s activities came under scrutiny from Meta as well. In its latest “Adversarial Threat Report,” Meta reported taking down several accounts and pages associated with Stoic due to suspicious activity. It highlighted the importance of combating AI misuse and collaborating with industry partners to ensure secure platforms for users.
OpenAI emphasized the need to invest in technology and teams to identify and disrupt actors engaging in malicious activities. The company stated that it utilizes AI tools to combat abuses and protect users from harmful influences. Despite the detailed report, OpenAI declined to provide further specifics on their disruption methods, prompting further inquiries from outlets like Dark Reading.
In this era of evolving technology and AI-driven operations, it is essential for organizations to remain vigilant and proactive in detecting and preventing malicious activities that seek to manipulate and deceive users. OpenAI’s efforts are a testament to the ongoing battle against misuse of AI for harmful intentions, emphasizing the importance of collaboration and innovation in maintaining a safe and secure online environment.