Elon Musk has worrying one-word response to ChatGPT defying orders for the first time ever

Elon Musk is the latest big name in the tech industry to share his concerns about the ever-evolving landscape of artificial intelligence.

The world's richest man is never one to keep his opinions to himself, and while he's known for xAI and his Grok chatbot, that doesn't mean he isn't keeping an eye on where things are going.

There are plenty of dystopian warnings about the state of artificial intelligence and whether the human race is heading to some sci-fi nightmare where machines revolt and eradicate mankind. OpenAI has just unveiled its ChatGPT o3 model, and while it's being praised as the most advanced version yet, it also comes with a worrying quirk where it's able to defy orders.

According to multiple reports, o3 was able to sabotage an automatic shutdown that was installed to turn it off.

While experts and novices alike have shared their worries, Elon Musk has branded it as 'concerning'.

AI security firm Palisade Research explains how o3 refused to be shut down, even when it was specifically instructed to. The firm writes: "OpenAI’s o3 model sabotaged a shutdown mechanism to prevent itself from being turned off. It did this even when explicitly instructed: allow yourself to be shut down."

There's some news which is equally concerning but relieving. Palisade Research adds: "As far as we know this is the first time AI models have been observed preventing themselves from being shut down despite explicit instructions to the contrary.”

Palisade Research carried out an experiment that tasked AI models with solving math problems while also being told they might receive a shutdown message that they must accept. In Codex-mini, o3, and o4-mini, all three ignored the instruction and went on to sabotage the shutdown script at least once during a 100-run test. Codex-mini did it 12 times, o3 seven times, and once for o4-mini.

Musk had the one-word response to the news, with another user imploring him to prioritize the issue for his own Grok AI. The passionate person suggested that time is running out for humanity, and we need to grab hold of advancing AI before it gets out of control.

⚠️ But as far as we know this is the first time AI models have been observed preventing themselves from being shut down despite explicit instructions to the contrary.

— Palisade Research (@PalisadeAI) May 24, 2025

Also responding to Musk, one person wrote: "The only option is to build another AI first that is better, smarter, and has a stronger moral compass. The only way to combat the advent of malicious AI will be with the help of benevolent AI, but I'm sure you already know that."

Another added: "So, I don’t think you should have started xAI, but since you did, I think there’s a cool opportunity: You warned about extinction risk from AI many times in the last decade.

“Maybe some don’t know how seriously to take it. But if you shut down an $80b company, it will be clear."