When the people building the most powerful technology in human history start buying the silence of those meant to govern it, we should all be paying attention. Most of us aren’t.
John O’Farrell left Andreessen Horowitz in May. He was the firm’s first outside general partner, hired in 2010. He left publicly, and he left angry. In a New York Times essay, he accused his former colleagues of spending nine figures to defeat politicians who want AI regulated and install those who won’t interfere. His conclusion was blunt: this will fail, and the backlash will be fierce.
He’s right. The problem runs deeper than lobbying.
Before the first regulation was ever written, before any politician was ever approached, the ethical foundation of AI was already being laid. It was laid by the people who labeled the data. Workers, often in Kenya, the Philippines, Venezuela, paid fractional wages to classify images, moderate traumatic content, and annotate the raw material that made large language models possible. The moral architecture of AI does not begin in boardrooms or Senate committees. It begins there, in that labor, mostly invisible, largely uncompensated at any meaningful level.
If you build on a broken foundation and then buy the people who might have corrected it, you are not innovating. You are just consolidating.
Power this concentrated, this unaccountable, never holds. Something always gives. The only question is whether the correction is designed or catastrophic.
What is actually being built here is not just technology. It is a new architecture of power. A small number of people with extraordinary capital are using AI to accelerate their own influence while simultaneously funding efforts to prevent democratic oversight. This is not a conspiracy theory. O’Farrell named it plainly. The mechanism is political action committees, candidate selection, and the strategic elimination of regulation before it can form.
The most dangerous part is not the technology itself. The most dangerous part is what it does to the population living alongside it.
Instant gratification is a business model. Every system optimized for immediate reward, for the frictionless answer, for the dopamine of having your question resolved before you finish asking it, is a system quietly eroding the cognitive muscle that makes citizens hard to manipulate. Critical thinking is not a natural state. It is a practice, built through resistance, through uncertainty, through the experience of not knowing something and having to sit with that long enough to figure it out.
A population that stops thinking critically is a population that stops asking who benefits.
That is extraordinarily convenient for the people who benefit most from not being asked.
This is not an argument against AI. It is an argument about intent. The technology itself is neutral in the way that any tool is neutral. What is not neutral is the decision to optimize for engagement over empowerment, for dependency over capability, for scale over equity. Those are choices. They are made by specific people with specific interests. And right now, those people are also funding the political infrastructure to ensure no one can hold them accountable for those choices.
O’Farrell believes a backlash is building. He is probably correct. Concentrated, unaccountable power has a historical tendency to generate its own opposition, usually at the worst possible moment, when the correction is messy and the damage already done. The pattern is not unique to technology. It is the pattern of every industry that moved faster than governance, convinced itself that regulation was the enemy of progress, and discovered too late that the absence of guardrails was not freedom but fragility.
Something always gives. The question is whether it gives by design or by crisis.
I
0 Comments