OpenAI dismantles two influence operations related to Russia and Iran

OpenAI has neutralized two distinct influence operations, "Dark Clark" and "Bogus Bylines," which utilized ChatGPT to generate and refine disinformation. The first campaign, attributed to Russian actors, focused on Latin America with the goal of denigrating Ukraine and regional leaders through fake audio on TikTok and deceptive outreach to Peruvian schools. The second campaign, attributed to Iranian actors, targeted the geopolitical tension between Washington and Tehran by placing nearly 100 fake articles in over a dozen media outlets using fictional Western journalist personas.
These operations signal a shift in disinformation tactics, moving from the proliferation of fake social media accounts toward the infiltration of established media outlets to increase reach and credibility. While OpenAI has provided specific impact ratings and actor attributions, the company acknowledges that some claims made by the operators remain unverified.

* OpenAI identified two disinformation campaigns using ChatGPT.
* One operation, "Dark Clark," is attributed to Russian actors.
* "Dark Clark" targeted Latin American countries to denigrate Ukraine and specific Latin American leaders.
* Actions included distributing a fake audio recording of the Ukrainian consul in Ecuador on TikTok.
* Operators used a fake email address to contact Peruvian schools regarding Ukraine.
* OpenAI rated "Dark Clark" level 5 out of 6 on its impact scale.
* A second operation, "Bogus Bylines," is attributed to Iranian actors.
* "Bogus Bylines" published fake articles regarding the conflict between the United States and Iran.
* ChatGPT was used to refine texts, meet editorial standards, and draft submission emails.
* Seven fictional authors posing as Western journalists signed the content.
* Nearly 100 articles were published in over a dozen media outlets since July 2025.

The strongest version of this narrative is that AI has lowered the barrier to entry for high-fidelity disinformation, allowing state actors to bypass traditional linguistic and editorial hurdles to infiltrate legitimate media pipelines. This represents a transition from "quantity" (bot farms) to "quality" (editorial mimicry).
The narrative relies heavily on the Authority Game. The central claims—the attribution of these campaigns to Russia and Iran and the "level 5" impact rating—rest entirely on the internal findings of a single private entity. There is no independent verification or third-party forensic data provided to support these conclusions.
The root cause is the fragility of the "editorial gatekeeper" model. The assumption is that media outlets can distinguish between a legitimate Western journalist and an AI-refined persona. This echoes historical patterns of intelligence agency "active measures," now supercharged by generative AI.
The implication is a potential erosion of trust in all digital journalism. If fictional personas can successfully seed 100 articles across a dozen outlets, the cost of verifying human identity becomes a systemic burden. The benefit goes to the AI provider, who positions itself as both the tool of the problem and the primary sentinel against it.
Bridge Questions:
How do we verify the attribution of these campaigns without relying solely on the vendor's internal telemetry?
If established media outlets were deceived by fictional authors, what does that reveal about the current state of journalistic vetting?
Counterstrike Scan:
A coordinated campaign pushing this narrative would seek to frame AI companies as the sole defenders of truth against "foreign bots" to justify increased corporate surveillance of AI usage. The content here is primarily a corporate disclosure and does not match the structural intensity of a coordinated attack.
Patterns detected: ARC-0043 Authority Game