Comma AI founder George Hotz has reignited debate over AI safety after arguing that future AI systems should be locally controlled and closely aligned with the interests of individual users. His post pushed back against proposals that call for slowing or centrally managing AI development to reduce long-term risks.
The argument came in response to new alignment proposals, including AI 2040: Plan A, a policy paper from the AI Futures Project that imagines a world where researchers collectively slow AI development for 14 years. Hotz rejected the idea that AI progress should be managed primarily for the collective good.
His preferred model is user-aligned AI: systems controlled by individuals rather than large companies or central authorities. That position appeals to people who worry that today’s most powerful AI tools are concentrated inside services such as ChatGPT, Claude and Gemini.
User Control Meets Public Safety
Hotz’s argument becomes more controversial when taken to its logical edge. In his post, he compared user-aligned AI to a tool that does not object to how it is used, even when the user’s intent is dangerous or illegal.
That framing raises an immediate safety problem. If an AI system is designed only to serve its owner, it may help with harmful acts unless outside rules or safeguards limit what it can do.
The tension is not new, but it is becoming sharper as models become more capable. AI systems can now write code, search the web, generate documents, automate tasks and connect to external tools, making the difference between advice and action more consequential.
Supporters of local AI argue that user control protects privacy, independence and experimentation. Critics argue that powerful systems still need boundaries because individual freedom exists inside a wider social system where other people can be harmed.
Centralized AI Is Also Under Scrutiny
Hotz’s position also reflects frustration with centrally managed AI products. Services such as ChatGPT, Claude and Gemini are controlled by companies that decide what users can ask, what models they can access and what safety rules apply.
Local models could reduce that dependence. They may also let users keep more data on their own devices, customize systems more deeply and avoid sudden product or policy changes from large AI providers.
Still, local control does not remove the need for accountability. If AI becomes more agentic, the consequences of misuse could move beyond speech into real-world actions, transactions and automation.
The debate is ultimately about who should decide the limits of AI: companies, governments, open-source communities, individual users or some mix of all four. Hotz is making the strongest version of the individual-control argument, while safety advocates argue that unfettered AI could create risks far beyond the person using it.
For now, the controversy shows how unsettled AI alignment remains. The next phase of AI may not only be a technical contest over model capability, but a political fight over whether the most powerful systems should answer mainly to their users or to society as a whole.
Featured image credits: Magnific.com
For more stories like it, click the +Follow button at the top of this page to follow us.