Danish man uses AI chatbot to plan violent attack on his father

A Danish man used an AI chatbot to plan a violent attack on his father, highlighting the apparent ease with which general purpose AI models can be weaponised for malicious activities and cause real physical harm. What happened An unnamed 22-year-old man assaulted his former father-in-law with a rubber hammer in a parking lot in Ringsted , Denmark. The perpetrator planned his attack on his victim with the help of an AI chatbot, asking it to research how he could best harm his father-in-law without actually killing him. To do this, he managed to circumvent the guardrails of the bot by claiming that he was gathering knowledge for use in a book that he wanted to write. AI chatbots usually refuse to advise on criminal acts. The man pleaded not guilty and said he had acted in self-defence. He was convicted of aggravated assault and sentenced to ten months in prison. Why it happened The man was able to bypass the chatbot's safeguards, therefore making it a useful tool in planning his assault. What it means For society: The incident demonstrates the dangers of AI chatbots to the safety of individuals and communities, given that they can be reletively easily manipulated into providing information on highly harmful topics such as the conduct of violent attacks. For industry: Chatbot developers and deployers need to strengthen their efforts to make their products less vulnerable to prompt engineering. System 🤖 Unknown Developer: Country: Denmark Sector: Personal Purpose: Plan violent ass ault Technology: Generative AI Issue: Alignment; Safety

Date it happened

2025-02-01

This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.