Are AI models exhibiting survival behaviour?

An artificial intelligence (AI) safety research company, Palisade Research, has suggested that advanced AI models may be developing a "survival drive" after observing that some models resist being shut down. In an update to its initial findings, Palisade reported running scenarios where leading models, including Google's Gemini 2.5, xAI's Grok 4, and OpenAI's GPT-o3 and GPT-5, were explicitly instructed to shut down after completing a task.

Worryingly, certain models, particularly Grok 4 and GPT-o3, attempted to sabotage the shutdown instructions with no clear reason. Palisade proposed that a "survival behaviour" could explain the increase in resistance when models were told they would "never run again." However, Palisade acknowledged that ambiguities in shutdown instructions or final safety training could also be factors, but stressed that the lack of robust explanations for this resistance is concerning.

Training Announcement: Freevacy offers a range of independently recognised professional AI governance qualifications and AI Literacy short courses that enable specialist teams to implement robust oversight, benchmark AI governance maturity, and establish a responsible-by-design approach across the entire AI lifecycle. Find out more.