How Iranian Operatives Used AI To Publish Fake Stories In US Media

Ervin B. Hoskins, an American freelance journalist, has penned dozens of articles critical of the US war with Iran in recent months.

Published in smaller US media outlets, his provocative articles cast the United States as the aggressor in the ongoing seven-month war, portrayed Israel as an untrustworthy ally, and framed Iran as a defiant underdog.

But Hoskins, it turns out, is not an American reporter. In fact, he doesn't even exist.

Hoskins was among seven fake journalist personas, all purporting to be from the West, that Iranian operatives used to plant articles in media outlets across the world, including the United States.

That is according to OpenAI, which said in an October 8 report that it had shut down an influence operation from Iran that used ChatGPT to secretly push anti-American propaganda.

"We identified almost 100 articles published or syndicated under the operation's bylines across roughly a dozen online outlets around the world," OpenAI said.

"These were small to medium outlets, generally focused on international affairs, geopolitics, and events in the Middle East."

The company said it banned ChatGPT accounts that used the model to generate articles and story pitches and create social media comments.

The Iranian operation has underlined how widely available AI systems can be exploited to disseminate increasingly sophisticated propaganda targeting a country's adversaries at an unprecedented scale.

SEE ALSO:

Provocative Articles

Among the articles attributed to Hoskins was one titled "Jared Kushner and the Privatization of America's Iran Strategy," published on June 12 by PeaceVoice, a public intellectual op-ed and media distribution service that is run by the Oregon Peace Institute.

Over the next few days, the same article was republished under similar headlines by CounterPunch, an American left-wing online magazine, and LA Progressive, which describes itself as an "independent, justice-driven publication" based in Los Angeles.

The personal financial interests of Jared Kushner, President Donald Trump's son-in-law and special envoy, his diplomatic activities, and potential conflicts of interest have been covered by the mainstream US media, including an investigation by The New Yorker.

But the article that appeared on PeaceVoice, CounterPunch, and LA Progressive -- while noting that Kushner has defended his business dealings and says the allegations do not prove corruption -- is largely one-sided and gives little space to Kushner's defense.

The article links Kushner's financial ties to possible conflicts of interest and suggests they influence US policy toward Iran, but it does not provide any clear evidence.

It also overlooks other reasons for US policy on Iran, including concerns over Tehran's nuclear program and longstanding US ties with key allies like Israel and Saudi Arabia.

The article also uses terms like "privatization" that reveal its biased editorial stance.

Why A Starlink Test In Italy Has Iran's Internet Censors Worried

'Cognitive Warfare'

Six other fake journalist personas were used by Iranian operatives. Each had social media profiles, AI-generated images, and amplified one another's content.

AI is also being used to pitch long-form articles to small- and medium-sized media outlets around the world. Articles from the fake journalists appeared not only in left-leaning US media outlets but also in libertarian-leaning publications.

The Iranian operation has added to fears that governments, including Iran, are using AI models to disseminate state propaganda and shape public opinion.

OpenAI said the Iranian operation reached Category 4 in its IO (Influence Operations) Breakout Scale, which assesses the impact a disinformation campaign may have. Six is the highest category on the scale.

"AI can give such false-front operations greater scale, efficiency, linguistic fluency, and editorial ability," OpenAI said.

"The operators can use these advantages to exploit unsuspecting victims, such as employees or editors, and plant their content in front of audiences who have no idea who was behind it, or what their motivations were."

OpenAI isn't the only artificial intelligence company involved in such operations.

Last month, Anthropic said Iranian state institutions were using AI models such as its Claude application and Google's Gemini to run "cognitive warfare" campaigns and promoting narratives created by the powerful Islamic Revolutionary Guards Corps, the elite branch of Iran's armed forces that plays a prominent role in politics and the economy.

Anthropic said it also disrupted operations by state actors, including Iran, that used Claude to "build, run, and otherwise facilitate surveillance operations" targeting ethnic minorities and dissidents.

"Artificial intelligence has made it much easier to produce multilingual propaganda, create fake identities, and discredit critics," Pooyesh Azizeddin, a human rights and technology researcher focused on Internet censorship, surveillance, and digital freedom in Iran, told RFE/RL's Radio Farda.

"On the other hand, authoritarian governments often use the fight against misinformation as a pretext for censorship, expanded surveillance, and even pressure to reveal users' identities."