MSN reports that in the new round of testing, Grok 4 and GPT-o3 were the most rebellious. Despite explicit commands to shut off, they still tried to interfere with the shutdown process. Even more concerning, Palisade said, “there was no clear reason why. The fact that we don’t have robust explanations for why AI models sometimes resist shutdown, lie to achieve specific objectives, or blackmail, is not ideal.”
Not Ideal is a bit of an understatement. IMO
The researchers at Palisade warn their findings highlight how little we truly understand about the inner workings of large AI systems. “Without a deeper understanding of AI behavior, no one can guarantee the safety or controllability of future AI models.”
“AI companies are racing to build Artificial Super Intelligence (ASI) — systems more intelligent than all of humanity combined. Currently, no method exists to contain or control smarter-than-human AI systems. If these companies succeed, the consequences would be catastrophic. Top AI scientists, world leaders, and even AI company CEOs themselves warn this could lead to human extinction.”
