1. News
  2. Everyday tech
  3. ChatGPT · Tech Support
ChatGPT · Tech Support2 min read

ChatGPT told some users to email reporters for tech help

A report says chatbots pointed people with locked accounts and broken phones to reporters. A Tom's Guide test found no sign of it, but the advice still needs checking.

By the Dashline deskSource: Tom's Guide
ChatGPT told some users to email reporters for tech help
Photo: Tom's Guide

Some people who asked ChatGPT for tech help were told to email journalists. The tech site 404 Media says it has received dozens of such requests in recent months, many written by AI. Tom's Guide then tried three similar problems and was not sent to a reporter once.

What happened

404 Media co-founder Jason Koebler described the pattern in a Monday report. One email came from a man in Russia who lost 13 years of Instagram photos and videos after his account was banned. He had used up Meta's appeals process. He then asked ChatGPT for another option. The chatbot suggested contacting journalists, including Koebler, and drafted the email.

Other cases were similar. One person had an iPhone stuck in SOS mode amid Verizon service problems and emailed several newsrooms asking for an investigation. Another, who was struggling with an unreliable ChatGPT assistant named Captain NORBERTA, sent three AI-made PDFs: a press release, a personal statement and an evidence appendix.

When 404 Media asked people how they found the outlet, it learned that AI chatbots were repeatedly sending them there.

What the test found

Amanda Caswell, AI editor at Tom's Guide, recreated three real-world problems. The first was being locked out of Instagram. ChatGPT gave standard recovery steps, then walked her through them one at a time, much like a customer service agent.

The other two involved a new iPhone. One was an old phone stuck in SOS mode, which limits a phone to emergency calls. The other was a forgotten Apple account password that blocked setup. The chatbot explained what was going on and stayed on the problem. It could not get around Apple's security rules, though.

What is not known

The test does not show that the problem is gone. Caswell notes that OpenAI may have fixed it internally, but that is a guess. The material does not say how often chatbots give this kind of advice, or whether a different wording of the question would bring it back.

What to do

ChatGPT can be a reasonable first stop for phone or account trouble. The article's warning is that an AI can write a convincing email and a plausible plan without pointing to a real fix.

  • Give the chatbot details, including what you already tried.
  • Treat each suggestion as a lead to check, not an order.
  • Use the company's own support channels too.
  • Be wary of any plan that sends you to a newsroom.
The takeaway

Use ChatGPT as a first stop for tech trouble, but check its advice before acting and also contact the company's own support.

More articles

You should really read these.

Next step

Let’s talk