OpenAI says it banned two covert influence operations that used false identities and AI-generated material to spread geopolitical messaging: a Russia-origin operation it called Dark Clark and an Iran-origin operation called Bogus Bylines. The company’s October 8, 2026 report describes different tactics and impact assessments for each, while noting that some claims could not be independently corroborated.
What OpenAI reported
OpenAI described both operations as combining AI with older influence tactics, including deceptive personas and the placement of content through purportedly independent fronts. It said the activity originated in Russia and Iran, respectively, and that it banned accounts associated with both. These are OpenAI’s findings from its investigation; the report is not independent verification of every allegation.
As an Amazon Associate I earn from qualifying purchases.
OpenAI characterized the campaigns as using multiple tactics to pursue their goals: “Both operations were multi-dimensional, trying a range of different tactics to achieve their goals.”
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Dark Clark: a purported research platform and local political messaging
OpenAI said the Russia-origin operation, which it nicknamed Dark Clark, targeted countries across Latin America. Its activity appeared aimed at undermining Ukraine’s reputation, alongside messaging about local politics, particularly in Argentina and Bolivia.
#1 Best Overall
The operation used ChatGPT most often to draft internal reports. OpenAI also said it was used to create content and prepare reports about a purported research platform called the Social Research Center (SRC). According to the company, operators appeared to control the SRC through the false persona “Mia Clark.” People working on the ground in Latin America appeared unaware of the Russian operation, OpenAI said.
OpenAI assessed Dark Clark at Category 5 on the Brookings IO Breakout Scale, which the company described as ranging from Category 1, the lowest, to Category 6, the highest. OpenAI called it the first Category 5 operation it had disrupted since it began publishing reports.
On claims related to reported 2024 campaigns targeting Argentina’s president, OpenAI wrote: “Both these claims have been publicly attributed to ‘Politology’ and ‘La Compania.’” That statement concerns those two claims; it does not establish that the attribution itself proves who was responsible.
Bogus Bylines: articles under purported journalist identities
OpenAI said the Iran-origin operation, nicknamed Bogus Bylines, used seven purported journalist identities to pitch long-form articles, mainly about the US-Iran conflict. It also generated batches of social-media comments and internal reports. OpenAI identified almost 100 articles published or syndicated under the operation’s bylines across roughly a dozen online outlets.
The company assessed the article-publishing activity at Category 4 on the Brookings IO Breakout Scale, citing consistent publication across outlets. It separately rated the social-media replies at Category 2 because it found little evidence that they broke out or dominated replies. The two scores describe distinct distribution channels, not one overall score for the Iranian operation.
OpenAI said the replies usually represented a minority of comments on posts and did not appear to attract substantial engagement. It also found a flaw in the operators’ internal measurement: they counted views of posts before and after replies appeared, rather than views attributable to the replies. OpenAI said that approach exaggerated the replies’ apparent impact.
What the reach figures do—and do not—show
One outlet that published articles from Bogus Bylines had almost 2 million Facebook followers, almost 355,000 X followers, and more than 544,000 Instagram followers, according to OpenAI’s figures as of August 2026. Those are account totals for the outlet, not readership figures for the specific articles. They do not show how many followers saw or believed the operation’s content.
The report’s other scale figures also need to be read in context: almost 100 articles across roughly a dozen outlets refers to the Iranian operation’s placements, while the Category 5, 4, and 2 assessments refer to different operations or channels. OpenAI also said it had exposed 30 covert influence operations since early 2024.
Best Value
How strong is the evidence for the reported impact?
OpenAI said some Dark Clark claims were corroborated by public fact checks and official denials, and that matching content could be found online. Other claims could not be corroborated in open sources. The company said possible explanations included removal, private circulation, or that the material never spread.
OpenAI also cautioned that operators sometimes claimed credit for events and arguments that were not their work. Their internal accounts of impact therefore should not be treated as verified outcomes. The report’s channel-specific assessments describe OpenAI’s evaluation of the activity; they do not establish that every targeted audience encountered or accepted the messaging.
For Bogus Bylines, published or syndicated articles and relatively weak social replies are materially different forms of distribution. OpenAI’s report says some outlets carried the articles, but it does not establish that those outlets knowingly participated in a covert operation or knew the bylines were deceptive.
Recommended Free Tools
OpenAI’s assessment of AI’s role
OpenAI’s broader assessment is that AI can help make false-front activity more scalable and polished, while public disclosure may make covert operations harder to continue. The report presents this as the company’s analysis, not as proof that AI alone caused either operation’s reach or impact.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
