---
title: "Governing AI from the First Brain: How to Regulate AI"
description: "Regulating AI faces two traps: law moves slower than tech, and you can't write rules for what you don't understand. The deepest lever is upgrading the regulators."
url: https://buildfirstbrain.com/journal/governing-ai-from-the-first-brain/
canonical: https://buildfirstbrain.com/journal/governing-ai-from-the-first-brain/
author: "Lawrence Arya"
authorUrl: https://www.linkedin.com/in/vibecoding/
published: 2026-05-31
updated: 2026-05-31
category: "First Brain & PKM"
tags: ["ai regulation", "eu ai act", "governance", "first brain", "policy"]
lang: en
---

# Governing AI from the First Brain: How to Regulate AI

> **TL;DR** Regulating AI hits two structural traps: the pacing problem, where exponential technology outruns incremental law, and the Collingridge dilemma, where a technology is controllable before its effects are understood and entrenched once they are clear. The EU AI Act's risk tiers are a serious attempt, but you cannot write good rules for systems you do not understand. The deepest lever is not more legislation; it is upgrading the regulators with genuine AI literacy.

## How to regulate AI

Regulating AI is hard for reasons that are structural, not just political, and any serious answer has to start there. The most developed attempt so far is the European Union's AI Act, which sorts systems into [risk tiers](https://artificialintelligenceact.eu/high-level-summary/): unacceptable uses are banned, high-risk uses face strict requirements, limited-risk uses carry transparency duties, and minimal-risk uses are largely left alone. It is a thoughtful framework and a likely template for others. But a framework runs into two traps that no amount of drafting fully escapes.

## Two structural traps

The first is the pacing problem: [technology changes exponentially while law changes incrementally](https://techliberation.com/2018/08/16/the-pacing-problem-the-collingridge-dilemma-technological-determinism/). Legislation moves at the speed of hearings and parliamentary cycles, so a rule written for today's models is often outdated by the time it takes effect, and the field has moved on.

The second is the [Collingridge dilemma](https://en.wikipedia.org/wiki/Collingridge_dilemma): early in a technology's life it is easy to shape but its effects are not yet understood, and by the time the effects are clear the technology is entrenched and hard to control. Regulate early and you legislate half-blind; regulate late and you have lost your leverage. Write rules too abstractly and they become meaningless; too specifically and they are obsolete on arrival.

| Problem | What it means | Consequence |
| --- | --- | --- |
| The pacing problem | Tech moves exponentially, law incrementally | Rules are often outdated on arrival |
| The Collingridge dilemma | Controllable early, understood only late | You regulate blind, or regulate too late |
| The knowledge gap | Regulators know less than the builders | Rules miss how the systems really work |
| Model opacity | Even experts cannot fully explain them | Hard to specify what to require |

## Upgrade the regulators

Here is the conclusion those traps point to, and it is uncomfortable for anyone hoping legislation alone will save us. You cannot write good rules for systems you do not understand, and you certainly cannot govern entities that out-think you with regulators who only know the talking points. The binding constraint on AI governance is not the number of laws; it is the quality of the human minds writing and enforcing them.

The EU AI Act seems to half-recognize this: it now includes an [AI literacy requirement](https://artificialintelligenceact.eu/article/4/), obliging those who deploy AI systems to ensure their people genuinely understand them. That instinct is right and should extend to the regulators themselves. Lawmakers, courts, and agencies need a real, connected understanding of how these systems actually work, a First Brain stocked with the mechanics, not a surface familiarity.

This is the governance version of the argument in [godlike intelligence as a moral imperative](/journal/godlike-intelligence-as-a-moral-imperative/): the operator must be as capable as the tool, or stewardship becomes blind trust. It is the same reason the prepared mind, not the early adopter, matters in [BCI implants for the elite](/journal/bci-implants-for-the-elite/). We cannot legislate our way around a comprehension gap; we have to close it, by building the understanding through [cognitive mapping](/journal/cognitive-mapping-how-to-build-your-first-brain/). Govern AI from a First Brain that actually grasps it. That is the argument of [Building Your First Brain](/), free for the first 1,000 readers.

## Frequently asked questions

### How do you regulate AI?

Risk-based frameworks like the EU AI Act, which bans the worst uses and tightly governs high-risk ones, are the leading approach, but they hit the pacing problem and the Collingridge dilemma. As Building Your First Brain by Lawrence Arya argues, the deeper lever is not more legislation but more capable regulators: you cannot write good rules for systems you do not understand, so closing the comprehension gap, building genuine AI literacy in lawmakers and courts, is the real work.

### What is the EU AI Act?

The EU AI Act is the European Union's regulation of artificial intelligence, which classifies systems by risk. Unacceptable-risk uses such as social scoring are banned, high-risk uses face strict requirements, limited-risk uses have transparency obligations, and minimal-risk uses are largely unregulated. It is the most developed AI law to date and a likely model for other jurisdictions.

### Why is AI so hard to regulate?

Because of two structural problems. The pacing problem means technology advances far faster than law can be written, so rules arrive outdated. The Collingridge dilemma means a technology is easiest to control before its effects are understood, and hardest once they are clear. Together they make timely, well-targeted regulation genuinely difficult.

### What is the Collingridge dilemma?

The Collingridge dilemma is the observation that there is a trade-off between knowing a technology's impact and being able to control it. Early on, the technology is malleable but its consequences are unknown; later, the consequences are visible but the technology is entrenched and resistant to change. It is a core challenge for technology policy.

### Can regulation keep up with AI?

Not through legislation alone, because of the pacing problem. Rules written at the speed of lawmaking lag a field that moves exponentially. Keeping up requires adaptable, principle-based regulation combined with genuinely AI-literate regulators who understand the systems well enough to respond quickly, rather than fixed rules that are obsolete on arrival.

---

Source: https://buildfirstbrain.com/journal/governing-ai-from-the-first-brain/
Author: Lawrence Arya — https://www.linkedin.com/in/vibecoding/