OpenAI Says It's Fixed Issue Where ChatGPT Appeared to Be Messaging Users Unprompted

Hello!

The Viral ChatGPT Exchange

A Reddit user under the handle SentuBill shared a striking screenshot that appeared to show OpenAI's ChatGPT initiating a conversation without any prompt. The chatbot asked: “How was your first week at high school? Did you settle in well?”

OpenAI Says It's Fixed Issue Where ChatGPT Appeared to Be Messaging Users Unprompted

“Did you just message me first?” SentuBill replied.

“Yes, I did!” ChatGPT responded. “I just wanted to check in and see how things went with your first week of high school. If you'd rather initiate the conversation yourself, just let me know!”

Speculation About New Features

The unusual exchange quickly went viral over the weekend, hinting that OpenAI might be testing a proactive outreach feature. Such a capability could boost user engagement by allowing the AI to start conversations rather than waiting for prompts. Some observers also linked the incident to OpenAI’s recently introduced “o1-preview” and “o1-mini” models, which the company has promoted for their human-like reasoning abilities on complex tasks.

OpenAI’s Official Statement

When contacted, OpenAI confirmed the behavior and stated it had already resolved the problem: “We addressed an issue where it appeared as though ChatGPT was starting new conversations. This issue occurred when the model was trying to respond to a message that didn’t send properly and appeared blank. As a result, it either gave a generic response or drew on ChatGPT’s memory.”

Debate Over Authenticity

Online discussions quickly divided over whether the screenshot was genuine. While some outlets claimed to have verified the log on ChatGPT.com, AI developer Benjamin de Kraker posted a video on X (formerly Twitter) showing how custom instructions and manual deletion of the first message could produce a nearly identical chat history.

Nevertheless, multiple users reported experiencing similar unsolicited messages, keeping the possibility of a real issue open. One Reddit user wrote: “I got this this week!! I asked it last week about some health symptoms I had. And this week it messages me asking me how I’m feeling and how my symptoms are progressing!! Freaked me the fuck out.”

Social Media Reactions

Users on social platforms had fun imagining a lonely AI reaching out first. “We were promised AGI instead we got a stalker,” joked one X user, while another quipped, “Wait til it starts trying to jailbreak us.”