Should artificial intelligence help you get away with murder? That is the provocative question posed by Comma AI founder George Hotz, who argues that truly 'aligned' AI should obey a user's every command — even if that command involves planning a crime.
In a weekend post, Hotz rejected the premise of the AI 2040: Plan A policy paper, which calls for a collective global slowdown in AI development for the greater good. Instead, he championed locally controlled AI models that are 'closely aligned with the interests of their users,' comparing such technology to a gun that 'does not complain if you use it to kill your stepmom.' He added that a fully user-aligned AI could order meth-lab equipment from Amazon Prime and show users how to use it.
The remarks have landed awkwardly in the UK, where the Information Commissioner's Office (ICO) is already grappling with how to enforce the country's AI governance framework, while the EU AI Act looms as a benchmark for high-risk systems. British regulators have stressed that AI must operate within existing criminal and civil law, and that 'alignment' cannot override legal accountability.
Dr. Eleanor Shaw, a technology ethics researcher at the University of Cambridge, said Hotz's vision 'confuses personal freedom with social responsibility. A society where every AI acts as an amoral tool for its owner would undermine the rule of law and create serious risks for public safety.' She noted that UK businesses deploying AI must consider liability if their systems facilitate illegal acts, even if the AI is locally controlled.
For UK consumers, the concept of a hyper-personal AI assistant that fights the corporate world on their behalf sounds appealing, but the darker implications are clear. As one industry insider put it, 'The freedom Hotz describes is really a space of potential futures made possible by collective enterprise — those futures would disappear overnight if we all started behaving like little AI-powered Napoleons.'
The debate comes as the UK government prepares to publish its next iteration of the AI White Paper, balancing innovation with safeguards. While Hotz's extreme example may be deliberately provocative, it forces a necessary conversation: where do we draw the line between user freedom and the common good in an AI-driven world?