Tech

‘Reading this at 2am is scary’: ChatGPT went off the rails—and people are freaking out

Users thought it was having a digital stroke.

Photo of Mikael Thalen

Mikael Thalen

Chatgpt with spanglish messages

ChatGPT unexpectedly began spouting off gibberish late last night, leading many concerned users to suggest that the AI has “gone rogue.”

Featured Video

In a series of posts to social media, users of OpenAI’s chatbot noted on Tuesday that the technology had started responding to their queries with incoherent remarks.

In one example, ChatGPT began speaking a garbled mix of Spanish and English, prompting the user to check in on the technology’s wellbeing.

“Are you having a stroke?” the user replied to the chatbot. “Some of what you’re saying makes no sense or aren’t proper words.”

Advertisement
https://twitter.com/seanw_m/status/1760115118690509168?s=20

In response, ChatGPT issued a three-paragraph-long apology that similarly descended into unintelligible rambling.

“!Whoops! I really apologize if my last response cam through as un unclear or se siente like it drifted into some nonsensical wording,” ChatGPT replied. “Sometimes, in the creative process of keeping the intertwined Spanglish vibrant, the cogs en la tecla might get a bit whimsical. Muchas gracis for your understanding, y Ill ensure we’re being as crystal-clear como l’eau from now on.”

The examples, which were primarily posted to the ChaptGPT subreddit, only got more bizarre from there. Another user who asked a music-related question saw the chatbot respond by repeatedly sending out musical note emojis alongside the phrase “Happy listening!”

Advertisement

Even requests as simple as asking the chatbot how to make sun-dried tomatoes produced strange answers.

“Forsake the new fruition morsel in your beloved cookery,” ChatGPT said.

One user even claimed to receive a somewhat intimidating response from the AI after asking it a coding question.

“Let’s keep the line as if AI in the room,” the chatbot wrote.

Advertisement

The user told fellow Reddit users: “Reading this at 2am is scary.”

The issue, which appears to affect version 4 of the chatbot, has been acknowledged by OpenAI.

In a statement on their website, OpenAI confirmed that the issue had been identified after a brief investigation. Although OpenAI did not explain what the issue was, but said it was “continuing to monitor the situation.”

Advertisement
web_crawlr
We crawl the web so you don’t have to.
Sign up for the Daily Dot newsletter to get the best and worst of the internet in your inbox every day.
Sign up now for free
 
The Daily Dot