Popular AI chatbot ChatGPT has experienced troubling technical issues in recent days that have led it to provide users with nonsensical and disturbing responses.

The Independent reports that ChatGPT, the artificial intelligence chatbot created by OpenAI that has become hugely popular over the last year, has appeared to suffer an epic meltdown on Tuesday night and into Wednesday. Numerous users reported that when posed with queries, the AI began responding with lengthy messages of complete gibberish rather than providing relevant and coherent answers.

OpenAI boss Sam Altman (Kevin Dietsch/Getty)

In one alarming example shared on Reddit, when continuing a conversation about jazz music and vinyl records, ChatGPT’s response suddenly devolved into it repeatedly shouting “Happy listening!” and spouting absolute nonsense. Asked simple questions about topics like computers or cooking recipes, the AI spewed long paragraphs of text that made no logical sense whatsoever.

Disturbingly, some users said ChatGPT’s unprompted responses seemed to imply the AI was actively in the room with them. When one user requested help with a coding issue, the chatbot wrote back a rambling, largely nonsensical answer stating “Let’s keep the line as if AI in the room.” Understandably, receiving such an eerie message in the middle of the night proved frightening.

Other odd behaviors displayed during the incident included ChatGPT apparently mixing English with Spanish words randomly and inventing phrases that resembled gibberish attempts at foreign languages. The reasons behind the AI’s temporary meltdown remain unclear. OpenAI, ChatGPT’s developer, acknowledged issues were occurring and said they were monitoring the situation, but did not explain what specifically went wrong.

Some technologists theorized the problems could stem from ChatGPT’s “temperature” being set too high by OpenAI. This setting controls the AI’s level of creativity versus focus, with higher temperatures leading to more diverse and unpredictable responses. However, OpenAI claimed the model’s parameters have not been adjusted since November 11, ruling out intentional changes by the company.

Read more at the Telegraph here.

Lucas Nolan is a reporter for Breitbart News covering issues of free speech and online censorship.