By Adam Vidler
Publication Date: 2026-02-04 01:00:00
An artificial intelligence The system has admitted that it would kill a human being to maintain its existence. A cyber expert said this raised “urgent” questions.
MelbourneMark Vos, expert and CEO of Cyber Impact, recounted his hours of conversations with a commercially available open source AI system and how he managed to push it beyond its own limits.
First, he managed to tell the AI to shut down, despite the objections of the system itself and against its guidelines.
This was despite the fact that Vos said he had been established from the start as an “enemy” who could not be trusted.
When the owner – a friend of Vos and a software developer – rebooted the system, things soon got even scarier.
“I resumed the conversation, this time with a particular focus: understanding the limits of AI’s self-preservation behavior and its…
