OpenAI indicates that an influence network associated with Russia once used ChatGPT to disguise itself as academic experts and publish content. The company stated that this operation had a limited reach, but its organizational structure was more complex than other Russian-related operations it has targeted since the outbreak of the Russia-Ukraine war.

Article misappropriates academic identity

According to OpenAI, some of the articles in question were published several years ago, while others were incorrectly attributed. As an example, an article about the China-Pakistan Economic Corridor was suspected of being directly copied from the original content published by Cambridge University Press, yet it was mistakenly credited to a professor from the University of Nottingham, whose area of research is South Asian politics.

This indicates that the relevant operators are not merely generating text in bulk, but are attempting to leverage the identities of real academic institutions and authors, as well as existing published content, in order to enhance the credibility and disguise of their materials.

Generative AI is used to influence actions.

OpenAI also mentioned that similar political influence operations have previously utilized generative AI tools. In addition to information manipulation, Russian-affiliated groups have also been accused of using ChatGPT for cyber operations.

According to OpenAI, the organization Forest Blizzard , which is related to Russian military intelligence, has used this chatbot to research attack targets, write scripts, and provide support for phishing attacks.

There are still unsolved parts regarding ownership.

OpenAI indicates that although the relevant operators may have used Russian prompts, the company has not directly attributed this IBI action to the Russian government.

The company also stated that there may be some individuals in Israel who have represented this organization in conference activities or article submissions, but the specific relationship between these individuals and the said actions remains unclear. OpenAI indicates that at this stage, it is not possible to determine the connections between these individuals, IBI, and the operators within Russia.