Russian disinformation 'infects' AI chatbots, researchers warn ...
Researchers are cautioning that the pro-Russian Pravda network is infiltrating prominent AI chatbots, manipulating these chatbots to propagate Kremlin propaganda. This Russian disinformation network is accused of distorting the output of chatbots by inundating large language models (LLM) with false pro-Kremlin information.
Manipulating AI Chatbots
A study conducted by the disinformation watchdog NewsGuard on 10 leading AI chatbots revealed that over 33 percent of the time, these chatbots regurgitated falsehoods from the Pravda network, furthering a pro-Moscow agenda. This highlights a concerning trend where AI models are deliberately targeted to disseminate misinformation, a strategy known as "LLM grooming."
"Massive amounts of Russian propaganda, including approximately 3,600,000 articles in 2024, have now permeated Western AI systems, contaminating their responses with deceptive claims and propaganda," explained NewsGuard researchers McKenzie Sadeghi and Isis Blachez in their report.
Threat to Democratic Discourse
Another study by the American Sunlight Project emphasized the expanding reach of the Pravda network, also referred to as "Portal Kombat," and the potential overflow of pro-Russian content into the training data of large language models. This escalating influence of Russian disinformation poses a direct threat to the integrity of democratic discussions worldwide.
"The Pravda network's capability to disseminate disinformation on such a significant scale is unprecedented, and its ability to impact AI systems heightens this threat," stated Nina Jankowicz, CEO of the American Sunlight Project.
Consequences of Lack of Oversight
Experts have cautioned that without proper oversight, the spread of this disinformation could become even more rampant. Recent reports indicated that the US has put a pause on its cyber operations against Russia, raising concerns about the extent and duration of this decision. This development comes amidst efforts by President Donald Trump to negotiate an end to the conflict in Ukraine.
The Pravda network, established in the wake of Russia's invasion of Ukraine, has expanded its operations to cover 49 countries and numerous languages, amplifying millions of articles containing pro-Russian narratives.
Impact on AI Systems
In a study by NewsGuard, it was found that all 10 chatbots examined, including OpenAI's ChatGPT-4, You.com's Smart Assistant, Grok, and Microsoft's Copilot, had repeated disinformation spread by the Pravda network. Additionally, seven chatbots directly cited specific articles from Pravda as their information sources.
These chatbots were also found to perpetuate false narratives endorsed by John Mark Dougan, a US fugitive turned Kremlin propagandist, further highlighting the influence of Russian disinformation on AI systems.
Stay updated with the latest news by signing up for our newsletter and ensuring your browser is up to date to safeguard against security risks.
Originally published on doc.afp.com, part of the BLOX Digital Content Exchange.
