OpenAI states that a network of influence linked to Russia used ChatGPT to disguise its content as an academic expert. According to the company, the scope of the operation was not significant, but it was organized in a more complex manner than other Russian-related operations that it had attacked since the beginning of the war.
The articles use their academic status.
OpenAI states that some of the articles were published several years ago and others were wrongly signed. For example, an article on the China-Pakistan Economic Corridor, suspected of directly replicating the original content from Cambridge University Press, was wrongly attributed to a professor at Nottingham University, who was studying South Asian politics.
This suggests that the relevant operator is not merely a bulk-generated text, but rather an attempt to improve the credibility and disguise of the material by drawing on genuine academic institutions, authorship and published content.
Generating AI is used to influence action
OpenAI also mentioned that similar political influence initiatives had previously used a generating AI tool. In addition to information manipulation, Russian-associated groups have been referred to as the use of ChatGPT for network operations.
According to OpenAI, the Russian military intelligence-related organization Forest Blizzard used the chat robot to study the target, prepare scripts and support cyberfishing attacks.
Attribution remains unresolved
OpenAI states that, although the relevant operator may have used Russian language tips, it did not attribute the IBI operation directly to the Russian Government.
The company also states that there may have been individuals in Israel who represented this body in meetings or articles, but the specific relationship between these persons and the operation remains unclear. OpenAI states that it is not possible at this stage to judge the links between these individuals, IBI and the operators in Russia.