AI bros say pump the brakes, while Washington says keep moving

If you run a small business, work at a church, teach school, manage a nonprofit, or simply use a computer every day, the argument happening inside the artificial intelligence industry may sound pretty far removed from your life.

It is not. The people building some of the most powerful AI systems in the world are openly arguing about whether they are moving too fast.

I have worked with technology since long before anybody was talking about artificial intelligence at the dinner table. I remember dial-up networking, building computers one machine at a time, installing business networks, and troubleshooting problems before Google could tell you the answer.

One lesson has followed me through all of it: technology usually moves faster than the rules around it. AI may be the biggest example yet.

Anthropic CEO Dario Amodei recently published an essay called “We Must Pace the Frontier.” His concern is that AI is increasingly helping engineers build the next generation of AI. That process, called recursive self-improvement, could accelerate development faster than researchers can understand and control what they are creating.

Amodei says that within six to twelve months, a sufficiently capable swarm of AI agents could potentially compromise huge portions of the internet.

That is a serious warning, but Amodei is not calling for everyone to shut down their computers.

His proposal includes independent third-party evaluators working inside major AI labs, common safety standards among companies in democratic countries, and eventually some form of international coordination. He also acknowledges a geopolitical problem: slowing American development too much while China continues forward could surrender the very lead that gives the US room to be cautious.

He is not alone in calling for stronger safeguards. OpenAI CEO Sam Altman has backed independent evaluators. Elon Musk said, “Dario is right.” Google DeepMind’s Demis Hassabis and Microsoft’s Satya Nadella have also supported the general direction.

President Donald Trump has pushed back against this slowdown argument. He has described some of the warnings as coming from “negative forces” and emphasized keeping the US ahead, saying that “whoever wins AI, wins.”

China, meanwhile, has called the pause proposal fear-mongering. At the same time, China is drafting its own mandatory safety rules aimed at preventing rogue AI agents from escaping human control.

That is an important detail. Even countries arguing against slowing development are talking about controlling the risks.

In the US, senators are considering a “duty of care” requirement that could allow the government to audit AI products and seek court action against models considered unsafe.

So this is not really a fight between people who want rules and people who want no rules. The disagreement is over how strong those rules should be, who sets them, and whether they slow development enough to cost the US its technological lead.

For people outside the AI industry, we can think of it in networking terms.

Suppose you are responsible for the computers at a church in Madison County. You want a firewall. You want good passwords, backups, user permissions, and protection against somebody accidentally clicking the wrong thing.

But you do not secure the network by ripping the cables out of the wall and turning off every computer.

Good security allows you to do useful work while reducing the chance that something goes terribly wrong.

That is how I look at AI.

I think there should be guidelines and rules. Independent testing makes sense. Companies developing systems this powerful should have responsibilities beyond simply getting the newest model to market first.

But for the most part, I do not think the US should deliberately hold itself back while we are ahead.

History gives us a rough analogy.

During World War II, the US knew that Germany was working on advanced weapons research. Imagine if our response had simply been, “This technology is dangerous, so we should slow down and be more careful,” while our opponent continued moving forward.

The Manhattan Project produced the atomic bomb, and the destructive consequences of nuclear weapons are impossible to dismiss.

But nuclear science did not stop with weapons. It also contributed to advances in medicine, energy, scientific research, imaging, and our understanding of the physical world.

The lesson I take from that history is not that we should recklessly race toward every new technology. It is that technological leadership matters when competitors are not standing still.

You do not win by pretending the race is not happening.

You move, you learn, and you build the guardrails as you go.

The comparison is imperfect. AI is not one secret government project. It is being developed simultaneously by private companies, governments, universities, researchers, and open-source communities around the world. That makes controlling it far more complicated.

It also makes a one-sided slowdown dangerous. Even Amodei recognizes that safety measures have to account for the American lead over China.

For ordinary people, all of this becomes much less theoretical when AI reaches the software you actually use.

These systems are already affecting bookkeeping, advertising, education, healthcare, cybersecurity, church administration, graphic design, programming, and customer service.

If you own a small business in Berea, you probably do not need to spend your evenings worrying about recursive self-improvement. You do need to know whether the software handling your information is secure.

Does your private data stay private? Can an AI agent send something, delete something, spend money, or make a business decision without enough human supervision? Who is responsible when it makes a mistake? Can you turn it off?

Those are practical questions, and they are the reason I do not dismiss the warnings coming from the people building these systems.

When the engineers building the machines start asking for independent inspectors, I listen.

When national leaders point out that technological leadership has economic and national security consequences, I listen to that too.

Those two positions do not have to be enemies.

The goal should be strong enough rules to catch dangerous behavior without building a regulatory wall so high that innovation simply moves somewhere else.

AI is going to keep moving. The question is whether we move with it intelligently.

Build the firewall. Make the backups. Set the permissions. Test the system.

But do not rip the cable out of the wall while the rest of the world stays connected.