At some point, maybe sooner than most people want to admit, there is going to be a war with artificial intelligence at the center of it.
Not the science fiction version. No giant robots marching down the street. But AI will be used to read the battlefield. It will be used to probe infrastructure, test the enemy's defenses, identify enemy combatants, guide weapons, and yes - even execute kill orders. AI doing anything a field commander could do. AI doing anything a general could do. But with omnipresent knowledge of the field of battle as well as the global implications of every move on the board - at light speed. The use of AI in warfare will extend beyond the battlefield and into our minds, advancing the science of manipulation and propaganda, shaping what entire populations see, fear and believe.
When that happens, the story we will be told will be simple. It is a fight between our AI - the good AI - and their AI - the evil AI. Good vs. Evil. Ours protects freedom. Theirs threatens civilization.
It is time for a reality check.
America is not China
A huge chunk of the AI community has gone off the rails - completely misrepresenting the current facts - and this is what motivated me to write this essay and produce the video.
The United States and China are not the same. There are real, serious differences, including differences of ethics and morality. America is not perfect. We have a concentration of wealth and power, surveillance, manipulation, corruption. I have called those failures out before and I will keep calling them out.
But America is not China. We do not operate a one-party surveillance state, even if both of our political parties edge us closer to that reality every day. We do not jail booksellers for stocking the wrong biography. We do not maintain labor camps for minorities - even if we must be mortified by our early history. We do not treat peaceful dissent as a threat to be erased.
The differences matter. Pretending they do not exist in the name of sounding balanced is not wisdom. It's not being fair or intellectual. It is simply dishonest.
The trap has a name
But being better than China, being freer than China, does not make America immune to the same temptation.
If a tool can deliver power, control it. If it can be weaponized, weaponize it, and do it now, before the other guy does.
This is an ancient trap and it has a name: Moloch. In the Hebrew Bible, Moloch was a power to which children were sacrificed. The God of Abraham and the Jewish People utterly rejected this. Today the name describes a race toward catastrophe in which every participant sees the danger clearly. We all understand what may lie ahead. But no one can stop running, because everyone is afraid the other side will get there first.
AI may be the most powerful version of this trap humanity has ever built.
Generosity as a strategy
The AI arms race is real. National security concerns are real and cannot be ignored. The United States cannot opt out while China, Russia, Iran, criminal networks, and terror groups build whatever they have the capacity to build. And remember that those suggesting they are only building for defense - are being disingenuous to put it lightly. In the world of AI, defense and offense are largely the same skill set. The model trained to find a weakness so it can be repaired can find it so it can be exploited. The system that detects propaganda can write it, brilliantly. Still, I am not arguing for disarmament. That would not be wisdom. It would be surrender. This too is an old argument that goes back to Moloch - if the good guys give up the guns, only the bad guys will have guns.
The reality check: I keep watching people describe China as the hero of open intelligence. The generous alternative to greedy corporate America and a controlling American government. In this framing, the United States created a problem and China arrived like the glittering armored knight on a white horse to democratize the solution.
China has been releasing remarkably cheap, remarkably capable AI models that anyone can download and run. The truth is - alternatives are good! Competition is good! But let's be clear about why China took this approach. Open models have helped its labs close the capability gap, attract developers, and build enormous ecosystems. They placed Chinese technology underneath products being built all over the world.
That is not generosity. That is strategy. You make the economics impossible to resist so that dependency grows. Then you hold all the leverage. We have seen this pattern from China before, in manufacturing, in steel, in electronics, in pharmaceuticals. It is a modern form of dumping. And where there is dumping, there is eventually anti-dumping.
On June 29th I wrote about this decades-old playbook: market capture first, dependency next, control later. On June 30th I made a prediction: you will know when China has a truly superior model, because that is the model that will not go open source. On July 7th, Reuters reported that Chinese authorities were meeting with major tech companies about restricting overseas access to those same open models.
I take no great pleasure in the prediction. Years of working with China simply made the ending obvious.
For some additional context - I have spent a good part of my career working with factories in China. I have been there several times and have always been amazed by the vibrancy of the culture, the art, the history. I have colleagues there I respect and consider friends. People can be friends. The developer working on AI in China probably wants exactly what I would want: to do the best work he or she can.
But China's leading labs do not exist outside the party state. They depend on Beijing's blessing for capital, for data, for the right to operate. When Beijing decides openness no longer serves China's interests, no researcher's ideals and no company's commercial preferences will change the outcome. The state will decide.
Maybe not so different here
When intelligence becomes strategic infrastructure, it gets locked down. We do not let people screw about with nukes either.
In June, the US government ordered Anthropic to deny its newest models, Fable and Mythos, to every foreign national, even foreign nationals inside their own organization. With no practical way to comply selectively, access was shut down for everyone. Around the same time, we learned OpenAI's latest model was also being held back. Eventually both were released, but not exactly set free. Security-sensitive requests get blocked or routed to older models. Mythos remains tightly controlled. And no one other than the insiders really knows what, if anything, was changed before release.
The point is not whether every restriction was unreasonable. The most powerful models carry real risks. The point is that advanced intelligence is no longer being treated as ordinary software. It is being treated as national security infrastructure, with access lists, oversight, and boundaries.
This does not make the United States and China morally equivalent. The systems are not equivalent. The methods are not equivalent. The consequences of dissent are certainly not equivalent. But the strategic incentive is identical: control the strongest intelligence before the other side does. Moloch.
China has already shown us what centralized control looks like when authoritarianism is the operating system. AI will only advance it. America presents the harder test, and an unanswered question: can a free society defend itself and remain free? How often have we traded freedom for safety, been told it was temporary, and actually gotten the freedom back?
Does anyone believe the United States government is above using AI for mass surveillance of its own people? We know the answer. And it does not matter who is in the White House - right or left. Both want control of our reality.
The old propaganda poster that hung on the wall said the same thing to everyone. The AI version says exactly what you need to hear. It knows your fears, your resentments, your history, your pain. Be it China or the United States, the most effective control will not be a soldier in the street or a book on fire. It will be a search result that never appears or a model that gives one side of an argument while sounding perfectly natural. You will not feel oppressed. In fact, you will feel remarkably well informed. We will all feel very sure of ourselves even as the reality beneath our feet is rewritten.
And if a government can break a civilization's shared reality, it does not need to conquer it. It can sit back and watch the civilization dismantle itself.
Is freedom worth the risk?
I wish there were a clean answer. America must defend itself. China will do the same and pursue its best interests. Open models distribute power, but they also spread dangerous capabilities and strategic dependency. Closed models can be safer in some ways, but they concentrate power in a few governments and a few corporations, and those institutions will decide what everyone is allowed to know and do.
The answer cannot simply be more control or no control. It is not a binary. Maybe it begins with a harder question: what are we trying to save? If the answer is freedom, are we willing to take the risks required to protect it? Because there will be risks. We will need more models, more labs, genuine open access. And we will need people who can think critically, who want to push back, and who will challenge the system - and be encouraged to challenge the system.
The world has never been a particularly safe place, and freedom has never promised safety. But a world where only governments and a few giant companies possess advanced intelligence does not sound safe either. Unless you are willing to be safe in a locked cage.
We are going to have to decide, and soon. Is freedom worth the risk?
Maybe the answer comes back to our ability to evolve. We must. And if we can evolve faster than the machines we are creating, so that those machines are trained not on what we were yesterday but on the best of what we could be - that is our chance.
The next war is already being fought, and it is being fought with AI. It is not good AI versus bad AI. It is just us. AI is the mirror. When we send it to war, it will not reveal the machine's soul.
It will reveal ours.
