Iranian campaign planted fake articles in real U.S. publications using ChatGPT
An article on Iranian operatives using ChatGPT to generate op-eds for small U.S. outlets prompts broader concerns about how cheap, scalable AI tools supercharge longstanding propaganda and influence campaigns. Commenters argue that many states and private actors already manipulate media narratives, and that the real issues are weak editorial safeguards, opaque platform policies, and public over‑trust in online content. Some see open models and critical media literacy as preferable to centralized controls, while others worry that both human and AI systems are increasingly saturated with state-backed disinformation from all sides.
Overall View on Propaganda and Online Content
- Many argue that most online content is designed to sell ideas or products, often dishonestly, and that AI simply amplifies existing trends in manipulation.
- Several note that fake news, sock puppets, and influence campaigns long predate AI; LLMs mainly lower cost and increase scale.
- Some see the article as important for calling this out; others see it as fearmongering timed to politics and focused on an official “enemy” state.
LLMs as Influence Tools and Their Behavior
- LLMs are described as text generators indifferent to truth, optimized to tell users what they want to hear; “hallucinations” are framed as inherent, not exceptional.
- One anecdote: ChatGPT confidently “found” a specific tree at a local nursery, with a realistic product page and stock count, but the nursery had never carried it. This is seen as dystopian and erodes trust in tools.
- Debate on whether such errors stem from models inventing details vs. underlying SEO tricks or site behavior feeding bad signals.
Trust, Verification, and Journalism
- Commenters highlight how easy it is to place polished opinion pieces in marginal or even mainstream outlets, often via PR and political machinery.
- There’s concern that smaller publications—blogs, local papers, opinion sites—are especially vulnerable to AI-generated op-eds masquerading as grassroots views.
- Others point out this has “always been the game,” referencing long-standing media manipulation, intelligence operations, and PR “submarining.”
US vs Foreign Propaganda
- Some stress that focusing on Iranian or Russian propaganda ignores extensive propaganda from the US and allies; accusations of “red scare” mentality appear.
- Others respond that propaganda can be true or false; what matters is intent and framing, not factual accuracy.
Possible Responses and Open Questions
- Suggestions include: better verification practices, reputation tracking, cryptographic signing of content, and critical media literacy (treat all media as potentially hostile).
- Debate over open-weight models: one view is that they prevent a single company or country from controlling who gets censored; another asks how this helps limit covert propaganda at all.
- Concerns remain about training data “poisoning” and the lack of robust safeguards at model and platform scale.