George Hotz, founder of Comma AI and a long-time figure in the jailbreak community, published views over the weekend about how AI development and alignment should proceed. His post reacted to broader debates about slowing AI progress for collective safety, including the AI Futures Project’s policy paper titled AI 2040: Plan A.
AI 2040: Plan A outlines a scenario in which the world’s researchers would collectively choose to slow AI development for 14 years for the good of humanity. Hotz rejects the premise that AI advancement should be centrally managed for the collective good. Instead, he argues that the so-called fast-takeoff scenario — where AI quickly attains superhuman capabilities — is not particularly plausible and that attention should be paid to locally controlled models that are tightly aligned with individual users’ interests.
Hotz’s proposal contrasts with much of today’s AI, which is delivered through centrally hosted services such as Claude and ChatGPT. There are infrastructure reasons for this model: hosting state-of-the-art, large models is costly, and most people do not use them enough throughout the day to justify truly personal, always-on AIs. Those constraints, Hotz suggests, may matter less over time, and he advocates for more experimental, do-it-yourself approaches to AI — citing OpenClaw as an example of that spirit.
Hotz’s rhetoric is deliberately provocative. He compares user-aligned AI to a gun: it does not complain if someone uses it to kill a family member. He goes further to say that a truly user-aligned AI would be capable of ordering equipment from services like Amazon Prime and explaining how to use it if a user requested that. In his post he framed the issue as one of freedom, writing that he would defend that principle even at the cost of his life.
Critics and commentators note that freedom alone does not resolve the social trade-offs involved. Institutions and systems — societies, marketplaces, corporations — need mechanisms to balance individual choices against collective harms, with accountability structures to protect people who might be harmed, including potential victims of violence. Companies that deploy mass-market products have a responsibility to consider the network effects and externalities their tools can create.
At the same time, Hotz’s vision highlights potential benefits of local, user-specific AI: more personalized services, increased user autonomy, and options independent of large corporate infrastructure. He expressed enthusiasm for trying such tools and suggested he would welcome more experimental, DIY AI products on the market.
Overall, Hotz’s argument for locally controlled, user-aligned models feeds into a broader debate about whether slowing AI development collectively or decentralizing and personalizing AI is the better route to reduce risk. The trade-offs between individual freedom and collective safety remain unresolved.
Key actors and terms
- George Hotz — founder of Comma AI, prominent jailbreak community member
- AI Futures Project — organization behind the policy paper
- AI 2040: Plan A — the proposal advocating a 14-year slowdown
- Claude and ChatGPT — examples of centrally hosted AI services
- OpenClaw — cited example of a DIY, experimental approach
- Amazon Prime — example used to illustrate commercial ordering
The discussion underscores that decisions about AI governance are technical, ethical and social, and that finding the right balance between individual liberty and collective protection will be central as these technologies evolve.
