Before there was artificial intelligence, there was the original kind. It is reading this sentence right now.
Every model we have built learned what it knows from us. A few hundred years of human writing, arguing, sketching, and working problems out, fed into a machine that learned the patterns in it. We talk about AI like it arrived from the future. It did not. It is a mirror held up to the back catalog of human thought. The word artificial is doing quiet work in that name. The first intelligence, the real one, is the person. Call it Original Intelligence. OI came before AI, and AI is built out of what OI already made.
Worth remembering right now, because it is easy to feel small next to a thing that writes faster than you and never sleeps. But speed was never the part that mattered most about a mind.
A person can choose. A model only runs.
A processor runs its instructions. A person decides. You can look at two roads and choose, and you can choose the wrong one on purpose, knowing it is wrong. That sounds like a flaw. It is the foundation everything else stands on. Because you can choose, you can be answerable for the choice. Free will and accountability are two sides of one coin. You cannot have the second without the first.
Nearly everything we have built as a society rests on that coin. Strip the law down to the studs and it is a very old, very refined system for a single job. Holding people accountable for what they do with their freedom. Every statute assumes a someone behind the act. A person who chose, who could have chosen otherwise, and who can be made to answer. Courts, contracts, consequences, all of it runs on that one assumption.
AI doesn’t fit the chair we built
Now put AI into that machine and watch it jam.
AI is a new kind of actor. It writes the contract, sends the email, approves the loan, moves the money. Real actions with real consequences. But there is no someone behind the act, not the way the law needs one. You cannot punish a model. It has no intent, no conscience, nothing it is afraid to lose. This is not a small philosophical wrinkle. Liability is now the most contested question in AI law, and the reason is structural. Our rules for fault were built for physical products with clear defects, and for human beings with intent. A system that gives a different answer every time it runs, with no person forming intent at the moment it acts, slips right between the two. When the model made the call, who chose? Often the honest answer is nobody did, and our whole idea of responsibility has nothing to hold.
The world is scrambling to write the missing rules in real time. Europe moved first. The EU AI Act became law on August 1, 2024, the first comprehensive AI statute anywhere, and it switches on in stages: the banned uses became illegal in February 2025, the rules for general models in August 2025, the heavy requirements in August 2026, with fines reaching 35 million euros or 7 percent of global revenue. The United States has no comparable federal law at all. Instead there are more than forty states writing their own, hundreds of competing bills, and the hardest questions left to the slow grind of the courts. The rules are coming. They are just coming after the thing they are meant to govern.
We named them frontier models
We do not call the leading systems the best ones, or the biggest. We call them frontier models. The four companies building them started a group in July 2023 and named it the Frontier Model Forum. The term comes from economics, where the frontier is the outer edge of what is currently possible. A fair description. Maybe a more honest one than they meant. Because a frontier is also, always, the place the law has not reached yet.
We have run this exact play before, with ground instead of intelligence.
Look at the Indian Territory in the early 1870s, the land that became Oklahoma. It was not lawless because of the people in it. It was lawless by jurisdiction. The tribal courts there could only try their own members, and no federal or state court reached across the line. So the Territory became a refuge. A person could act, then ride past the edge of any court’s authority, and that was that. The gap itself was the draw.
Then the gap closed, and you can put a date on it. On March 18, 1875, President Ulysses Grant appointed Judge Isaac Parker to the federal court at Fort Smith, on the Territory’s border, and handed that single court jurisdiction over all seventy-four thousand square miles of it. Parker was authorized to swear in two hundred deputy marshals to ride in and bring people back to answer. One of them, Bass Reeves, would bring in more than three thousand over his career. The same act that in 1874 meant a clean ride out, by 1876 meant a courtroom in Fort Smith. Parker did not change human nature. He closed the jurisdiction gap. That is all the law ever does on a frontier, and it always arrives after the people are already out there. Frontier is not a neutral word. It means the order is still weeks behind the edge of the map.
So that is where we are with AI. Out past where the rules have been written and enforced, in the lag, with the marshals still being sworn in.
Who is actually steering
Here is where I land, and it is not the obvious place.
In that gap, the question that matters is not who you can blame when something breaks. A person to point at after the fact is worth very little. The question is who is steering while you are out there. A frontier was never crossed well by people who showed up with a wagon and good intentions. It was crossed by the ones who knew the ground, read the weather, and got the whole party through.
Same with AI right now. Pointing a powerful model at your business and hoping is the wagon-and-good-intentions plan. The value is in the hand on the reins. Someone who knows the tool well enough to catch it when it is confidently wrong, who has seen the specific ways these systems fail and steers around them before they cost you anything, and who carries an operation all the way to done instead of to an impressive demo. Casual AI use will not do that. Casual systems-and-automation knowledge will not either. It takes experience, a real appetite for making a business run cleaner, and genuine skill with the tool.
That is what Original Intelligence is for in this stretch. Not a name to pin the blame on. A hand that knows what it is doing, standing behind real work.
The frontier always comes before the law. While we wait for it to catch up, the edge does not belong to whoever has the model. Everyone has the model. It belongs to the person who can actually wield it.