Last week I read something that stuck with me more than the usual campaign noise. On October 8, CNN reported that OpenAI had caught a group of operatives using its models to write and place opinion pieces critical of the American war with Iran. They prompted ChatGPT in Farsi, revised drafts and editorial pitches with it, and, according to OpenAI, invented seven journalists who got roughly 100 articles into about a dozen outlets around the world. NPR and The Washington Post covered it too.
None of the outlets were household names. One was LA Progressive, a small Los Angeles site that ran a piece titled “Is America's Iran Policy Being Privatized? The Jared Kushner Question.” Another was Middle East Monitor in London, which has close to two million Facebook followers. After the pieces ran, the operators went back to ChatGPT and asked it to write comments that would look like ordinary readers reacting to them.
OpenAI banned the accounts. It could not say who was behind the work, and it did not tie it to the Iranian government. It said the activity looked like a commercial actor running an influence campaign for hire. We are used to imagining propaganda as a ministry somewhere. This sounds more like a small business with a service menu.
My first reaction was the obvious one. This is a con. Somebody pretended to be people who do not exist, and editors trusted them. Small publications run on freelance pitches and thin staff, and a polished draft from a plausible name is exactly what they hope to find in the inbox. AI makes that draft cheap to produce, in clean English, at volume. Jessica Brandt, who ran the federal government's Foreign Malign Influence Center from 2023 to 2025, told CNN it was remarkable that the material reached real outlets, because that is what extends its reach. A post from an anonymous account is easy to ignore. An article with a byline in a real publication borrows that publication's credibility.
But I have a second reaction, and I think it is worth saying out loud. The arguments in those pieces were not alien ideas smuggled into American life. Plenty of Americans oppose this war. Plenty think the gas prices and the threat to shipping through the Strait of Hormuz are too high a cost. Plenty of others believe the war is necessary and that Iran has earned the pressure. Both camps are real, and both include people writing in good faith. The problem with this operation is not the opinion. It is the disguise. Readers deserved to know that the “journalist” was a fiction paid for by someone with a stake in the outcome.
I worry that stories like this get used in two bad ways. One is to shrug, since the outlets were small and OpenAI itself tends to be cautious about impact. In August it described a separate Russian operation built around a fake Israel-based think tank and said it “appears to have reached relatively small audiences.” But that same report made a point I keep returning to: the significance lay less in the audience than in the infrastructure, the fake experts and fake rankings that could lend credibility later. Small today can be a test run for tomorrow.
The other bad use is to treat every critic of a war as a possible foreign plant. That would be its own kind of damage. A healthy country argues about whether to fight. If the lesson people take from this is that dissent is suspect, the operators win something bigger than a few clicks.
So where does that leave me? Mostly with a plain appreciation for boring verification. An editor calling a writer before publishing. A video chat with a new contributor. A byline that connects to a real history. These are old habits, and they cost time that small newsrooms rarely have. Still, they are the cheapest defense available, and they work no matter which side the fake writer is arguing.
There is also an odd note of hope in the story. These operators got caught partly because they used a commercial AI tool that keeps records and has people watching for this kind of abuse. In its August report, OpenAI noted that the Russian operators' own use of AI helped expose them. That will not hold forever, as open models spread and anyone can run them privately. But for now, the same technology that makes the forgery cheap also leaves fingerprints.
As a reader, I am trying to adopt one small habit. When a piece makes me feel strongly, I look at who wrote it and whether that person seems to exist anywhere else. It is not a cure, but it is something.
Thoughts? Reach me at danielk@outlook.com.