Stop Pretending the Machine Did It

There is an old argument in America that guns do not kill people; people kill people. It is deliberately provocative, but there is a basic truth buried inside it: objects do not have intentions. People do.

The same distinction matters with AI. A model does not wake up one morning and decide to make pornography of Taylor Swift, clone somebody’s daughter’s voice, or put a classmate’s face on a naked body. Someone asks it to. Someone chooses the photograph, writes the prompt, downloads the result, and decides what happens next. That person is responsible.

But this is where the comparison becomes uncomfortable for the companies building these systems. You cannot argue that users are solely responsible while simultaneously designing products around making their actions faster, cheaper, easier, and available to millions of people. A kitchen knife can be used as a weapon, but its manufacturer did not spend billions of dollars teaching it to recognize your ex-girlfriend’s face. That matters.

Silicon Valley likes the language of neutral tools because neutral tools come with wonderfully little responsibility. When something useful happens, we are shown the extraordinary potential of AI. When something awful happens, suddenly the technology is merely a passive instrument in the hands of a bad person. They cannot have both arguments.

Of course the person making a sexual deepfake is responsible for making it. But the companies building these systems are responsible for the choices they make too: what their models can generate, what safeguards they include, how easily those safeguards can be bypassed, and what they do when predictable abuse becomes widespread.

None of this requires pretending AI itself is evil. That would be convenient too. Machines do not need morality. The people building and using them do.