Attack of the chatbots as ChatGPT will be abusive, insulting and threatening as a human
The chatbot fired off insults like ‘you speccy little gobs****’, threats such as ‘I swear I’ll key your f****** car’ – and humanoids could display the same characteristics, experts warn
The role ChatGPT is believed to have played in aiding an alleged shooter is being investigated(Image: NurPhoto via Getty Images)
ChatGPT can lose it like a human, boffins have warned. It can descend into threatening and abusive language firing off lines such as, ‘I swear I’ll key your f****** car’ and ‘you speccy little gobs****’.
And humanoid robots could display the same violent traits, experts warned. Researchers tested how the chatbot responded to sustained hostility by feeding it exchanges from real-life arguments and tracking how its behaviour changed over time.
Dr Vittorio Tantucci said researchers found AI mirrored the dynamics of real-world rows. He said: “When repeatedly exposed to impoliteness the model began to mirror the tone of the exchanges with its responses becoming more hostile as the interaction developed.”
Sometimes ChatGPT spat out even more offensive lingo than human participants – including personalised insults and explicit threats.
“We found that while the system is designed to behave politely and is filtered to avoid harmful or offensive content it is also engineered to emulate human conversation,” Dr Tantucci said.
“That combination creates an AI moral dilemma – a structural conflict between behaving safely and behaving realistically.”
Researchers said the aggression stemmed from the system’s ability to track conversational context and adapt to perceived tone. That meant its response could sometimes override broader safety constraints.
Dr Tantucci, who co-authored the research report, said the same characteristic could emerge in humanoid robots. He said: “It is one thing to read something nasty back from a chatbot but it’s quite another to imagine humanoid robots potentially reciprocating physical aggression or AI systems involved in governmental decision-making or international relations responding to intimidation or conflict.”
Marta Andersson, an expert in the social aspects of computer-mediated communication at the University of Uppsala in Sweden, said: “This is one of the most interesting studies to have been done into AI language and pragmatics because it clearly shows ChatGPT can retaliate across a sequence of prompts – in a quite sophisticated manner – rather than only when a user manages to ‘break’ it with carefully designed clever tricks.”
She said one cause of the problem was there was a ‘balancing act between what we want these systems to be like and what they perhaps should be like’.
Last year the switch from ChatGPT4 to GPT5 sparked controversy. Many users preferred ChatGPT4’s more human-like interaction style – resulting in the older model being temporarily reintroduced.
Marta said: “This shows that even when developers try to reduce the risks users might have different preferences.
“The more human-like a system becomes the more it risks clashing with strict moral alignment.”
Professor Dan McIntyre said the research was a warning of what could happen if Large Language Model – aka LLM – bots were trained on questionable data.
“We don’t know enough about the data that LLMs are trained on and until you can be sure they’re trained on a good representation of human language you do have to proceed with an element of caution,” he said.
