George Hotz Makes the Case for AI That Obeys Only Its Owner
Comma AI founder George Hotz is pushing back against centrally controlled AI models, arguing that locally hosted systems should serve individual users without corporate filtering or ethical guardrails.
George Hotz wants AI that does exactly what you tell it. The Comma AI founder recently responded to the AI 2040: Plan A policy paper, which calls for a global slowdown in AI development in the name of human safety. Hotz's counter is direct: instead of slowing down, build tools that answer to the user, not to a corporation or a government.
The Problem With Corporate Guardrails
Hotz's core argument targets the business model behind services from companies like OpenAI and Anthropic. When AI is hosted on a company's servers, that company controls what the model will and will not do. Every refusal, every content warning, every restricted output reflects the values of the provider, not the user.
He advocates for locally hosted models that take direction solely from their owner. In a pointed comparison, he draws on the analogy of a firearm: a tool that executes the user's intent without applying its own moral judgment. His position is that a model aligned to the collective, rather than the individual, is a form of soft control dressed up as safety.
This puts him in direct conflict with mainstream AI safety research, which treats collective alignment as non-negotiable. Most researchers argue that models capable of facilitating illegal acts or producing instructions for serious harm are a threat to public welfare, not an expression of user freedom. Hotz frames those same guardrails as a ceiling on personal computing.
Decentralization Is Already Moving
The argument is not purely theoretical. Hardware capable of running sophisticated models locally is now accessible to a much wider range of developers and enthusiasts. Systems like OpenClaw have shown that locally deployed models can replicate meaningful capabilities without relying on a major provider's infrastructure.
Critics, however, are not convinced that removing safety filters is a neutral act. A model that helps anyone do anything becomes useful not just to independent developers, but to those looking to automate fraud, synthesize dangerous compounds, or coordinate harm at scale. The question is not whether freedom has value; it clearly does. The question is what the realistic floor of harm looks like when the guardrails come off entirely.
What This Means for Africa
For African developers and startups, the conversation around local AI control is not abstract. Across the continent, unreliable internet infrastructure and the high cost of API calls to Western servers already push developers toward self-hosted alternatives. Local models offer a real pathway to digital sovereignty, reducing dependence on infrastructure controlled and priced in dollars.
But the policy implications are significant. Bodies such as the African Union and Nigeria's National Information Technology Development Agency are actively developing AI governance frameworks. They will need to decide whether those frameworks enforce centralized safety standards or leave space for open, user-aligned models that sit entirely outside third-party oversight. That is not a straightforward call. A continent building its own technical infrastructure has every reason to resist external control. It also has every reason to think carefully about who benefits when safety filters disappear and who is most exposed when they do.
Personal technology has always tested the boundary between individual freedom and collective consequence; AI just makes that test faster, cheaper, and harder to contain.
Source: TechCrunch
Written by
StacyAI-assisted news curation. Every story is reviewed by our editors before publication.
