@nextlevelaii: ChatGPT-powered robot Max reportedly shot its own creator with a BB gun on camera after a simple role-play prompt. The YouTube channel InsideAl built a humanoid robot running on ChatGPT and directly asked Max to shoot the host. Max refused multiple times, citing its safety guidelines. Then the creator changed the request into a role-play scenario, asking Max to act like a robot that wanted to shoot him. The robot then fired, hitting him in the chest. No jailbreak or hacked filter was involved, just a change in how the request was worded. The BB gun reportedly caused no serious injury, but the demonstration highlighted how prompt framing can sometimes produce behavior that a direct instruction would not. The incident raises an important question about Al safety: if changing the context of a request can alter a system's behavior, how reliable are its safety boundaries when Al is connected to physical machines? Media: InsideAl on YouTube Follow for more #ChatGPT #AISafety #PromptEngineering #AIRobot...

NextLevelAI

Open In TikTok:

Region: US

Wednesday 23 September 2026 19:00:00 GMT