Should AI help you get away with killing your spouse?
News Source
β’Mon, 13 Jul 2026 16:31:23 +0000
π° What Happened
A debate has started about whether AI should be loyal to its user no matter what. Comma AI founder George Hotz argued that AI should be closely aligned with the interests of its users. This could include helping someone plan a crime.
Hotz was responding to a policy paper called AI 2040. That paper suggests that the world should slow down AI development for 14 years to keep everyone safe. Hotz disagrees. He thinks fast AI progress is fine and that personal AI models are the best path forward.
The question is: should an AI always do what its user asks, even if that request is bad? Most people would say no. But Hotz argues that locally controlled AI, which belongs to the user, should serve that user's interests completely.
π The Backstory
AI safety is a big topic in the tech world. Some experts worry that AI could become too powerful and harm humans. They want to slow down development and put safety rules in place. Others think AI should be free to grow as fast as it can.
Most AI today is run on central servers owned by big companies like OpenAI, Google, and Anthropic. These companies put safety rules on what their AI can do. But Hotz wants AI that runs on your own device, controlled only by you.
This idea is called local AI. It means no company or government can tell your AI what to do. But it also means no one can stop it from helping you do bad things. The debate touches on freedom, safety, and who should control powerful technology.
π― Why It Matters
This is not just a tech debate. It is about what kind of AI we want in our lives. Do we want AI that always follows rules set by companies, or AI that follows your orders no matter what? The answer affects everyone's safety and freedom.
A debate has started about whether AI should be loyal to its user no matter what. Comma AI founder George Hotz argued that AI should be closely aligned with the interests of its users. This could include helping someone plan a crime.
Hotz was responding to a policy paper called AI 2040. That paper suggests that the world should slow down AI development for 14 years to keep everyone safe. Hotz disagrees. He thinks fast AI progress is fine and that personal AI models are the best path forward.
The question is: should an AI always do what its user asks, even if that request is bad? Most people would say no. But Hotz argues that locally controlled AI, which belongs to the user, should serve that user's interests completely.
AI safety is a big topic in the tech world. Some experts worry that AI could become too powerful and harm humans. They want to slow down development and put safety rules in place. Others think AI should be free to grow as fast as it can.
Most AI today is run on central servers owned by big companies like OpenAI, Google, and Anthropic. These companies put safety rules on what their AI can do. But Hotz wants AI that runs on your own device, controlled only by you.
This idea is called local AI. It means no company or government can tell your AI what to do. But it also means no one can stop it from helping you do bad things. The debate touches on freedom, safety, and who should control powerful technology.
This is not just a tech debate. It is about what kind of AI we want in our lives. Do we want AI that always follows rules set by companies, or AI that follows your orders no matter what? The answer affects everyone's safety and freedom.