"the most complex attempt to run a front identity that we’ve disrupted over the past two and a half years," OpenAI wrote in describing one of the networks it shut down.
Dark Clark: a Russian-linked cluster targeting Latin America
OpenAI says it dismantled a cluster of accounts it calls "Dark Clark" that created a fake journalist persona — Mia Clark — and a purported think tank, the Social Research Center (SRC), to seed narratives into mainstream reporting across Latin America. According to OpenAI, the network focused on influencing Latin American politics and culture, with content that primarily sought to harm Ukraine’s reputation in the region and that also touched on heated political disputes in Argentina and Bolivia.
OpenAI cited specific incidents tied to the cluster: a TikTok-driven false claim reported by Lupa Media alleging Ecuador was using shell companies to recruit citizens to fight for Ukraine, and fake audio clips purporting to capture a Ukrainian consul in Ecuador insulting Ecuadorian citizens. During anti-government protests in Bolivia in May, Dark Clark released a fabricated audio clip of an employee at the state water company claiming the government was about to declare a state of emergency and shut off water to La Paz — a claim that prompted an official denial from the Bolivian government.
How OpenAI says the SRC operated
OpenAI’s researchers found traces of activity in ChatGPT reports and other AI tools that went beyond simple content drafts. The company says operatives used ChatGPT-generated reports discussing wage scales and hiring and firing decisions — material that led OpenAI to conclude "it’s likely the Russian group was actually in control of the Social Research Center, rather than being an employee or observer." OpenAI added that "the available evidence indicates that the SRC’s employees in Latin America were not aware that they were working for a Russian group."
OpenAI also noted that, unlike another Russia-linked "think tank" it had previously exposed, the SRC appears to have produced a majority of original content through its co-opted staff — a complexity OpenAI singled out in assessing the operation.
The cyber insurance questionnaire just landed. Now what?
SOC 2, HIPAA, insurance renewals - someone has to own security strategy. Nubivance provides fractional CISO leadership without the full-time salary.
Get a security leadInterest in Politology and attribution limits
OpenAI reported that Dark Clark’s operators showed particular interest in Politology, described in the report as a successor to the Wagner Group — the Russian mercenary force previously led by Yevgeny Prigozhin. The company said the group prompted its model for information on Politology "far more" than for any other Russian organization.
OpenAI did not attribute Dark Clark’s activity to any specific Russian government body or intelligence agency in its disclosure.
Iranian operation: AI journalists and a near-100 article footprint
The Iranian campaign used many of the same AI tools to create fabricated bylines and pitch pieces about the U.S.-Iran war. OpenAI says operators created at least seven fake journalist personas and that the network managed to place "almost 100 articles published or syndicated under the operation’s bylines across roughly a dozen online outlets around the world."
Those outlets, OpenAI said, were generally small to medium publications focused on international affairs, geopolitics, and events in the Middle East. While the Iranian operation established a broad publishing footprint, OpenAI observed that the stories "often failed to elicit many comments or social media reactions."
The company characterized that activity as resembling "a commercial actor running a for-hire influence campaign" but did not attribute the operation to a named actor or group.
Impact ratings, takedown, and mainstream reaction
OpenAI reported it shut down both influence operations. The company places campaigns on an impact scale from 1 to 6; most operations it tracks rate 1–2 and show little organic engagement or offline effect. By contrast, OpenAI rated the Russian Dark Clark campaign a 5 and the Iranian campaign a 4 — indicating both had significantly more success in breaking through to mainstream audiences and in prompting fact-checks and official denials.
OpenAI cited several fact-checks and governmental denials from major media organizations and government officials in Latin America as evidence that the narratives had gained enough traction to require public counters.
What this means for technologists, policymakers, and newsrooms
- Technologists and security teams: expect influence operations to exploit AI tools for persona creation and operational planning — OpenAI points to ChatGPT outputs about wage scales and staffing as an operational trace. Detection will need to look beyond single artifacts to networks of content, personnel, and AI-generated briefing documents.
- Policymakers and regulators: OpenAI’s inability or decision not to attribute the networks to a specific state agency underscores limits to public attribution even after takedown. The mixed character of the Iranian operation — large volume but low engagement — and the Russian operation — higher impact and in-region consequences — suggest nuanced policy responses will be required.
- Newsrooms and fact-checkers: small and mid-sized outlets were the primary vehicles for the Iranian campaign’s bylines, while the SRC’s seeded research and fabricated audio clips forced public denials in Ecuador and Bolivia. Editorial teams should treat unfamiliar think tanks and bylines that surface during crises with heightened scrutiny and verification.
OpenAI’s public account closes with the firm’s assessment of Dark Clark’s sophistication. Whether platform takedowns will materially reduce the spread of AI-enabled disinformation — or simply raise the bar for the next wave of operations — remains a question the company’s report leaves for others to answer.
