OpenAI said on October 8, 2026, that it had banned a cluster of Russia-origin ChatGPT accounts linked to an influence operation it nicknamed Dark Clark. The company said the operation targeted audiences across Latin America and that its operators mainly used ChatGPT to draft and update internal reports; campaign-content generation was a minority of the activity OpenAI observed.

What OpenAI said about Dark Clark

OpenAI said the operation sought to weaken Ukraine’s reputation in Latin America and, in some cases, influence local politics, particularly in Argentina and Bolivia. The company’s report described activity involving fabricated documents and audio, as well as claims about school events in Peru and Ecuador.

OpenAI said the operators used VPNs to access ChatGPT because the service was not available to them from Russia. The company described internal reporting—not the creation of campaign material—as the principal observed use of the service in this operation.

ChatGPT’s role in the operation

According to OpenAI, operators used ChatGPT chiefly to draft and update internal reports about their activities, including reports sent to an unidentified superior. The company said it observed campaign-content generation in a minority of cases.

That distinction gives the report’s account a specific shape: ChatGPT was used mainly to document the operation, while some observed uses involved producing material for its campaigns. OpenAI’s findings concern the account activity it observed.

The purported research front and school-related claims

OpenAI described the Social Research Center (SRC) as a purported research platform managed through a fabricated persona named Mia Clark. The company said SRC staff interviewed regional experts and drafted research papers. It identified well over 60 articles on the SRC website, most of which appeared to be original compositions.

An SRC-branded LinkedIn account, as of August 17, 2026, showed 961 followers, claimed 200–500 staff, and listed two employees, OpenAI said. The company assessed that local employees appeared unaware they were working for a Russian operation and were doing research in good faith.

In Peru, OpenAI said operator reports described a fake email impersonating Lima’s Regional Directorate of Education and intended to prompt schools to hold Ukraine-related events. The company said it found public reporting that matched the operators’ account.

OpenAI also said operators claimed that a fake email prompted schools in Ecuador to hold a ceremony pledging allegiance to President Daniel Noboa and Erik Prince. The company’s account describes this as an operator claim; it does not establish that every detail in the operators’ reports occurred as they said.

OpenAI’s assessment and the limits of its account

OpenAI rated Dark Clark Category 5 on its IO Breakout Scale, which runs from 1 to 6. The company said it was the first Category 5 operation it had disrupted since it began publishing reports on such operations. The rating is OpenAI’s assessment under its own scale.

OpenAI said it could not corroborate some claims in the operators’ internal reports. It also said operators sometimes exaggerated their effectiveness or claimed credit for unrelated events.

Ecuador’s education minister responded to a school-related allegation

OpenAI cited a public rebuttal by Ecuadorian education minister Gilda Alcívar concerning the alleged school ceremony. Alcívar’s post is in Spanish and remains in its original form here.