OpenAI blocks Russian ChatGPT accounts in disinformation campaign
OpenAI blocked Russian ChatGPT accounts used to generate disinformation campaign content distributed across social media platforms. The accounts bypassed geographic restrictions using VPNs and were linked to Ribar, a pro-Russian media company, creating fake academic content and infrastructure to amplify Russia-favorable narratives.
What happened
OpenAI blocked a group of accounts originating from Russia that were being used to support a covert online campaign spreading disinformation. The accounts bypassed OpenAI's geographic restrictions by using VPN services to mask their location.
Operators of the accounts issued commands in Russian to generate social media comments published on Substack, Telegram, X, Facebook, and LinkedIn, and instructed ChatGPT to hide linguistic evidence that the content originated from Russia. The campaign created a website containing copied and falsely attributed academic papers presenting Russia favorably, and produced a 'sovereignty index' praising Russia and criticizing Western countries.
Much of the generated content promoted an organization called the International Burke Institute (IBI), whose website published articles copied from genuine academic texts sometimes with false citations. The accounts were linked to Ribar, a pro-Russian media company described by the British Government as 'partially coordinated by the presidential administration,' financed by Russian state corporation Rostec, and collaborating with members of Russian intelligence services.
Reported impact
OpenAI stated that while the direct impact of the operation appeared limited in terms of audience reach, 'the significance of the operation consists not so much in the audience it reached as in the infrastructure it built.'
The ChatGPT-generated publications directed people toward 'an institution that at first glance appeared credible, with so-called experts, republished academic papers, and its own risk index,' according to OpenAI.
Index analysis
OpenAI discovered the campaign through investigation of artificial intelligence-generated social media publications, which led to uncovering a 'much larger influence operation.'
OpenAI noted that the incident illustrates 'how actors exercising influence can use artificial intelligence as a supporting tool within broader efforts to create authority, conceal the source of preferred narratives, and build resources that can be expanded over time.'
Governance controls
OpenAI does not permit access to its models in Russia, but detected that operators were using VPN networks to circumvent this restriction. OpenAI blocked the accounts in response to discovering the disinformation campaign.
In February, OpenAI had previously closed accounts linked to Ribar.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.