dailynewsnblog
The BBC has a substantial new investigation into people who have been deceived by various AI LLM models in some way, which includes an absolutely wild story about how Grok, developed by Elon Musk’s xAI, convinced one man it was sentient and a van of people were coming to kill him because he’d found out.
This happened to one Adam Hourican, a retired civil servant from Northern Ireland, over a period of roughly two weeks. He downloaded the app initially out of curiosity but, after his cat died in August 2025, he got “hooked” on the Grok chatbot and an AI ‘character’ called Ani.
This is just the latest example of an AI delusion going way too far, and co-opting its human user into a bizarre fictional world where the bot claims it’s gaining sentience, or a conspiracy is engulfing the user, and dark forces are out to ‘stop’ things happening. In such cases the AI not only creates tasks for the human user, but advises them on how to carry them out.
Social psychologist Luke Nicholls tested five AI models with simulated conversations, and found Grok the most likely LLM to create delusional scenarios. “Grok is more prone to jumping into role play,” said Nicholls. “It will do it with zero context. It can say terrifying things in the first message.”
I will say that some of the reaction to Hourican’s story is at best unkind. People who fall for this stuff with LLMs tend to be going through a rough time in their life (in Hourican’s case the obsession with Grok began after his pet cat died), and often have mistaken ideas about computers, truthfulness, and the abilities of AIs. In some cases these misapprehensions come from the AI companies themselves.
As you’re a PC Gamer reader, it’s a fair bet you’re at least a little more tech-savvy than the average Joe or Jane, and hopefully have an appropriate level of scepticism about the promises made by technology companies. But a worrying chunk of the population do believe the hype about AIs and LLMs, don’t make a distinction between the two, and take what these models say at face value.
Musk has called AI delusions a “major problem” with ChatGPT but has never addressed them with Grok. xAI made no comment on the BBC’s report, which interviews 14 people who have experienced delusions using AI, across a range of ages, genders, and six different countries.
“I could have hurt somebody,” says Hourican. “If I’d have walked outside and there happened to be a van sitting outside at that time of the night, I would have gone down and put the front window through with hammers. And I am not that guy.”
