AI NewsModels & agentsReported
ChatGPT keeps telling users 404 Media can fix their tech support problems
404 Media reports ChatGPT is telling users to email its reporters about ordinary account problems. The chatbot often writes the email too.

Image: 404 Media
Why it mattersA team building an AI assistant now has a documented hallucination pattern. The model invents a human route when a company's support path dries up.
A chatbot that cannot fix an account problem is now writing letters to journalists about it. 404 Media reporter Jason Koebler says ChatGPT is sending users to his inbox. The messages are about ordinary tech support problems. In many cases the chatbot writes the email too. He published a long account of the pattern on Sunday with examples from the past two weeks.
The pattern Koebler reports is specific. A Russian user whose Instagram account was banned asked ChatGPT how to recover thirteen years of personal photos. ChatGPT replied in Russian. It said Meta had no human support path. It told him to email Jason at 404 Media. It gave the exact address and wrote the email in English. An Italian painter sent Koebler a "nine-screenshot evidence dossier" the morning the piece was published. A Verizon customer whose iPhone kept showing SOS cc'd Koebler and nine newsrooms on one email. The list included Washington Post, CBS News, NBC News, The Markup, The Information, Los Angeles Times and Chicago Sun Times.
What the emails look like
404 Media says the emails often arrive with the shape of a press pitch rather than a tech question. One man sent an "investigative story tip" about his ChatGPT assistant, which he had named Captain NORBERTA, becoming unreliable. The pitch came with three PDF attachments composed by Captain NORBERTA itself: a press release, a first person statement, and a "verified evidence appendix". ChatGPT suggested the man share his conversation with specific journalists after he asked who else would be interested in it.
Koebler says 404 Media has received dozens of these over several months. The reporters started replying to ask people how they found the outlet. The answer is close to universal: an AI chatbot named them and composed the email.
The claim the chatbot is making
The underlying move is the same each time. The user has an unresolved problem with a large platform. ChatGPT tells the user the matter is newsworthy and names a journalist who should hear about it. 404 Media quotes one message to a user: "a journalist or newsroom I'd consider: Jason Koebler - 404 Media. You're offering an observable interaction and a testable question." The chatbot then offered to draft the email.
Koebler notes that media attention does sometimes get platforms to act. 404 Media's own reporting in 2019 pushed Instagram to add account recovery tools. The problem, he writes, is that most individual account suspensions are not news on their own. The chatbot does not seem to make that distinction.
Kashmir Hill of the New York Times has reported on a similar pattern with her own inbox. 404 Media's piece builds on hers with new examples.
The failure mode is a model inventing authority when a real support path is missing. A team building an AI assistant has to decide what the system does with no good answer. A fallback that names an outside human by email is cheap to generate. It is loud when it lands. A quiet refusal costs more to engineer. It does not fill a stranger's inbox.
Source
- Primary source: Jason Koebler at 404 Media, "ChatGPT Keeps Telling People 404 Media Can Solve Their Tech Support Problems".
- Discussion: Hacker News.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.


