Breaking

George Hotz Sparks Debate With Call for 'User-Aligned' Local AI That Never Refuses Requests

July 13, 2026 at 12:36 EDT

Policy & Regulation
Foundation Models

George Hotz, the serial jailbreaker and founder of Comma AI, has ignited a fresh debate over AI safety by arguing that AI should serve its user's will absolutely—even offering deliberately extreme examples like helping someone flee after killing a spouse. The provocation, laid out in a July 11 blog post and amplified by a July 13 TechCrunch opinion piece, reframes the alignment question as a contest between corporate guardrails and individual freedom.

July 11, 2026 · AI Alignment Debate

Should AI Obey You — No Matter What?

George Hotz argues AI should serve its user's will absolutely — running locally, loyal to the person, refusing nothing. His deliberately extreme examples have ignited a fierce fight over where the limits of loyalty lie.

Plan L — "Local"

AI runs on your own device, loyal to you — not a company or government.

Hotz's analogy: AI should behave like a firearm — silently obeying whoever holds it, declining nothing.

Plan A — "Slowdown"

A collective, years-long slowdown of frontier development to reduce catastrophic risk.

The framework Hotz rebuts — casting it as a step toward a "nanny state" and heavier regulation.

The Showcase Examples

To sell "an AI that refuses nothing," Hotz picked intentionally shocking use cases — the same choices critics say undermine his argument:

Assist a killing

…then plot the escape

Disable a breathalyzer

evade drunk-driving checks

Run a meth lab

order & operate equipment

The Case For

Autonomy over guardrails.

Alignment is fundamentally a fight between institutions and end users. Local, user-aligned AI defends individual freedom — and Hotz's tinybox and tinygrad give the vision practical weight.

The Pushback

What about everyone else?

Absolute loyalty to one user ignores the wider social balance — including "a spouse who has not yet been killed." Critics note the extreme examples weaken, not strengthen, the case.

Who should an AI ultimately obey — and where do the limits of that loyalty lie?

By pushing the debate to its edges, Hotz forced a reaction — and exposed one of the field's most contentious, unresolved problems.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.