Iranian campaign planted fake articles in real U.S. publications using ChatGPT

Iranian campaign planted fake articles in real U.S. publications using ChatGPT(washingtonpost.com)

55 points by cainxinth 2 hours ago | 28 comments

slowin 27 minutes ago |

This is what Israel has been trying to do. I'm skeptical of US MSM pushing anti-Irainian propaganda, but you can see Israeli propaganda all over the internet in plain sight:

https://www.theguardian.com/world/2026/aug/26/fake-thinktank...

ghostDancer 5 minutes ago | |

Trying? Actively doing.

mschuster91 1 minute ago | |

You don't need propaganda to know that the Iranian Mullahs aren't your friends.

SirFatty 22 minutes ago | |

I'm always skeptical of random comments pulling Israel into the conversation that doesn't have anything to do with Israel.

foenix 21 minutes ago | | |

Explain how a war between US/Israel and Iran, that Israel started, has nothing to do with Israel?

orangedog 18 minutes ago | |

Should they just not report on it?

slowin 16 minutes ago | | |

Like how the New York Times "reported" on Iraq having weapons of mass distruction?

https://www.pbs.org/newshour/show/the-new-york-times-wmd-cov...

OgsyedIE 42 minutes ago |

Given the plurality of self-hosting options there is no reason to believe that it's not that case that every government on Earth is doing it to some extent.

Perhaps the real problem is too much acceptance of shoddy verification.

capitainenemo 31 minutes ago | |

There's also the other side of it, poisoning the AIs... https://www.newsguardtech.com/special-reports/moscow-based-g...

"Massive amounts of Russian propaganda — 3,600,000 articles in 2024 — are now incorporated in the outputs of Western AI systems, infecting their responses with false claims and propaganda."

in 2024.. hopefully filters when training have gotten better, but maybe not..

gustavus 38 minutes ago | |

I don't think it is a verification problem. It wasn't like before in the days of controlled media and radio there wasn't a narrative to push, it's just that most people didnt have any awareness of alternatives. Nowadays people are more aware of alternative viewpoints or perspectives. The solution isn't to try and force one approved source of truth, but to just assume everyone is trying to push their version of the truth and examine it critically.

My understanding is that this is the way France operates where you assume all the media is hostile and you view it all critically. In America this is a bit of a shock because for a long time we had trusted arbiters of truth and now that they are less trusted we've been trying to look for alternatives rather realizing it should all be examined critically.

sva_ 34 minutes ago |

https://archive.is/EFOhL

gizmodo59 33 minutes ago |

how is this news? Adversarial country using AI to generate fake news goes way before AI.

hammock 32 minutes ago | |

Why does it have to be an adversarial country? Why does it have to be a country at all? Couldn’t it be the mayor of some town? Or a PR team at a fast food franchise group? Or one of SBF’s polycule members?

The notion we only see this news when it is calling out a country with which the federal government is in conflict makes me question to veracity and intent of the article itself, more than the content of it.

deepsquirrelnet 20 minutes ago | | |

And what do you call it when a country acts in adversarial ways to its own citizens? Call it cynical, but it sure seems like there's a lot of that in the world right now.

bilbo0s 19 minutes ago | | |

Well it’s mostly other entities running influence campaigns.

Thats true.

But this article is about the influence campaigns presumably tun by the Iranian government.

bitmasher9 29 minutes ago | |

This is classic FUD, Fear Uncertainty and Doubt.

It’s designed to make the reader question which articles they have read recently that are Iranian propaganda, and sow doubt about future articles that seem to benefit Iran.

This amount of FUD always seems to increase around election time…

Telemakhos 17 minutes ago | | |

It's a well-researched article that lays out evidence, not suspicions. It includes OpenAI's own investigation into accounts that were found generating op-eds for Iranian propaganda. It shows the route by which these op-eds entered into smaller publications, namely by a pacifist NGO in Oregon, and neither then NGO nor the publications saw them as propaganda, despite the origins of the op-eds.

gustavus 47 minutes ago |

Remember 95% of what you see or hear on the Internet is designed to sell you something either an idea or a product, and it isn't done honestly.

everdrive 45 minutes ago | |

Much of modern internet content is far worse than cable advertising in the 90s. I remember when I was a kid; I'd aggressively mute the commercials because I hated the idea that someone was attempting to influence me. The modern internet is orders of magnitude worse buy nearly any measure. The nastiest, most outrageous content is plastered on every wall. Every single person has an angle. Every bit of content meant to pull you in one direction or another, and given the addictiveness of novelty and powerful emotions, we keep coming back for more. It's an absolute disaster we don't know how to walk away from.

bilbo0s 30 minutes ago | |

>95%???

That low?

In all honesty, given the volume of comments, posts and articles that bots can generate in a given time period? I mean it’s gotta be several orders of magnitude higher than what humans can generate.

And then you factor in the fact that even most of the humans are astroturfing..

The percentage of influence campaigns we interact with has just gotta be really high.

jadar 22 minutes ago |

democracy dies in the darkness of paywalls, I guess.

tyjen 46 minutes ago |

Not a bad actor spreading misinformation, but I encountered a similarish weird, dystopian experience with ChatGPT.

Utilized ChatGPT on my PC to search local plant nursery inventories for a specific tree. Output showed several local nurseries with rough probability descriptions regarding whether they have the tree in inventory, except for one nursery. For that nursery, ChatGPT provided a direct link to the specified tree, which was directly on the nursery's domain, showed a picture and description of the tree, and even showed the item in-stock with 6 in inventory at the 5-gallon size.

Excited, I plan my trip and am about to pick up my tree, I then double-check the nursery site on my phone to verify location and stock. As I attempted to navigate the website inventory to find the tree, I noticed that the tree no longer existed. There is no page for that varietal or species of tree on the entire website's domain.

I head back inside and look at the ChatGPT provided link on the PC and there it is, showing the inventory that doesn't show up when you use the nursery's website to directly find it without the provided link. A call to the nursery, further verified that the tree is not in stock, nor have they ever carried it.

ChatGPT made it all up. Did ChatGPT decide to do this, or did the nursery pay for this type of "advertising" to generate in-person customers? Nevertheless, still an incredibly weird and dystopian experience.

garyfirestorm 42 minutes ago | |

It didn’t generate a link (sources) when it claimed it was in stock

ButlerianJihad 17 minutes ago | |

Did you consider that the website made it up and fed it to ChatGPT?

It sounds like a bundle of very common SEO tricks combined with the differential between your human user agent’s visit.

Perhaps ChatGPT only showed you what the nursery told it about.

This phenomenon is not unique to LLM. If you telephone a store, chances are that the salesman who answers the phone will tell you only what you want to hear. They don’t even strictly need to lie. The store only wins if you drive there and walk in the door. Why should they discourage you from doing that very thing?

This concept was driven home to me at the dry cleaners. Why should a dry cleaner stop work and answer the phone, only to tell a customer that their batch isn’t ready? They just got taken off-task for a non-revenue interaction. It doesn’t matter, because even if the customer’s batch is ready, they aren’t there to pay and pick it up!

Avicebron 40 minutes ago | |

Not being able to trust your tools is certainly dystopian...