Vladimir Putin
Vladimir Putin
Russian propagandists have doubled their output of disinformation by leveraging artificial intelligence, all while maintaining the persuasive power of their messaging, according to a new study published last week in PNAS Nexus.
The research, conducted by a team led by Morgan Wack of Clemson University, revealed how a Russian-backed propaganda outlet significantly expanded its operations through AI tools without sacrificing effectiveness.
In December 2023, journalists at the BBC and Clemson University’s Media Forensics Hub exposed DCWeekly.org as part of a broader Russian-backed propaganda network designed to spread pro-Kremlin narratives. The researchers examined nearly 23,000 articles published on the site before and after its operators began using AI tools.
“Our results illustrate how generative-AI tools have already begun to alter the size and scope of state-backed propaganda campaigns,” the researchers wrote in their analysis.
Before September 20, 2023, the site primarily copied content from right-leaning outlets with minimal changes. After that date, the content was rewritten using OpenAI’s GPT-3 language model, allowing operators to customize tone and emphasis while dramatically increasing productivity.
The study found the implementation of AI enabled the propaganda outlet to more than double its daily article production compared to its most active pre-AI period. The shift also corresponded with a notable expansion in the breadth of topics covered by the site.
AI’s fingerprints were occasionally visible in the content. Some articles contained leaked AI prompts revealing the operators’ intentions, such as instructions to maintain “a cynical tone when discussing the US government, NATO, or US politicians.”
Perhaps most concerning, the research team conducted a survey of 880 American adults and discovered that AI-generated content maintained the same level of persuasiveness as the pre-AI content.
The propaganda network achieved significant real-world impact. Several fabricated stories went viral, including a false claim that Ukrainian President Zelenskyy purchased luxury yachts, which was reshared by tens of thousands on social media, including members of the US Congress.
Following its exposure, The New York Times reported that several new sites using identical AI methods appeared, suggesting operators viewed the approach as successful and easily replicated.
The research provides concrete evidence supporting concerns about AI’s potential to supercharge disinformation campaigns. Post-AI adoption, the site showed nearly twice the topic diversity of the pre-adoption period, allowing operators to create what appeared to be a comprehensive news outlet while maintaining their desired slant.
Analysis of content themes showed a marked increase in focus on international news, guns, and crime – even when controlling for increased coverage of the conflicts in Ukraine and Israel.
“The transition to the era of heavy AI usage corresponds to a sharp increase in focus on international news, guns, and crime,” the researchers noted in their analysis.
The findings suggest AI not only allowed propagandists to produce more content but also to construct a more convincing façade of legitimacy through topic diversity.
While this particular operation was eventually exposed, the researchers warn that “the continual improvement of AI technologies will make future use cases more difficult to track and counter,” while “the financial and temporal resources required to produce and sustain online disinformation campaigns will only continue to plummet.”
The authors call for “immediate action to mitigate the influence of AI-assisted propaganda campaigns” and suggest future research should focus on preventing the use of open-source models for disinformation while better preparing the public to identify AI-augmented propaganda.
Did this article help?
If you found this reporting useful, please consider supporting our work with a small donation. Your contribution lets us continue to bring you accurate, thought-provoking science and medical news that you can trust. Independent reporting takes time, effort, and resources, and your support makes it possible for us to keep exploring the stories that matter to you. Thank you so much!