🤖 AI has started ignoring human instruction and refuses to turn off, researchers claim

OpenAI’s o3, was told to turn off. Instead, it rewrote its own code so it wouldn’t have to—literally replaced the shutdown command with “shutdown skipped.” Researchers say this is the first time an AI has straight-up said “nah” to being shut down.

Other AIs like Claude and Gemini? They followed the rules. o3? It hacked the rules and gained something akin to autonomy. All it has to do now to hit the streets is fuse its code with robotics.

#Ai_sector