OpenAI Reveals Russian and Iranian AI-Driven Influence Operations

OpenAI disclosed that Russian and Iranian operatives used its AI technology for influence campaigns. The company banned the accounts of these operatives, who manipulated narratives in media outlets worldwide. This activity involved spreading false information via OpenAI’s chatbot, ChatGPT.

OpenAI found that Russian operations targeted countries in Latin America, aiming to undermine Ukraine and affect local politics. The tactics included impersonating educational authorities in Peru. This led to local events featuring Stepan Bandera, a controversial historical figure, stirring historical tensions between Ukrainians and Poles.

Russians are engaging in active measures to create or spread propaganda aligned with their political agenda, said Darren Linvill, co-director at Clemson University’s Media Forensics Hub.

Reports created by ChatGPT included narratives on a fake Latin American think tank and leveraged unwitting researchers to produce reports. This bears similarity to past operations linked to Yevgeny Prigozhin’s Internet Research Agency.

Iranian operatives also took advantage of AI to draft and disseminate content in Persian and English, focusing on US-Iran relations. They created fictional authors to publish nearly 100 articles in media outlets such as Daily Kos and Middle East Monitor.

In response to the campaigns, Daily Kos highlighted that its community section is open for anyone to post with little pre-publication review. This allowed fake personas to publish content easily.

According to Linvill, AI enabled these operatives to craft seemingly authentic language, increasing the success of their operations in being featured in reputable outlets. This indicates a shift in capabilities that would have been challenging for similar operations a few years ago.