The Constitution Lives and Breathes. AI Doesn't. Nobody Planned for That.

What happens when a living document meets something that has never been alive?

By DeLeigh Pearce Poole | The Poole Law Group, LLC

Thursday night, the federal government sent a letter to an AI company.  By Friday morning, a private product used by a good number of people no longer existed. No crime charged. No judge. No jury. No hearing. No evidence presented to anyone with the authority to say no.

Just a letter.  And it was done.

I've practiced criminal law for almost twenty years. I've watched the government do things that made me uncomfortable. I’ve seen the same government do things that made me cringe but that I understood. 

But I have to be honest with you — this one makes me question it all. 

---

The Rule That Built This Country

Here's what I need you to understand before anything else: before the government can take something from you — your freedom, your property, your livelihood — it has to prove something. To someone who can push back.  That principle is older than the republic. It is, in many ways, the republic.

It did not apply this weekend.

The Glock Problem

Let me put this in terms I actually work with every day.

Glock makes handguns in Smyrna, Georgia, about twenty minutes outside Atlanta. Every single day, somewhere in this country, someone uses one to commit a crime. 

Every single day that same product is used to protect someone else from a crime. And every single day, the government does the same thing: they go find the person who committed a crime. They charge that person. They prosecute that person. 

A gun doesn't make decisions. A person does. The manufacturers of Glock designed a product that could be used for good or ill.  

And in this country, before the government can take serious action against you, it has to establish two things: that you did something wrong, and that you meant to. Act and intent. Both of those have to live inside a human being.  Not the product.  Not in the creators of something that could be used either way. 

That's not a technicality. That's the whole architecture. The instrument is beside the point. 

Now ask yourself what happens when you take the person out of the equation entirely.

What Happens When There's Nobody to Charge

The government's stated reason for shutting everything down technically:

Fable 5 and Mythos 5 could be weaponized into tools for mass cyberattacks and surveillance.

If you haven't heard of them, here's the short version. Fable 5 and Mythos 5 were large-scale AI models built by Anthropic — one of the most prominent artificial intelligence companies in the world, founded by former OpenAI researchers and long considered one of the more safety-conscious players in the space. These weren't experimental tools or niche products. They were among the most capable AI systems ever made available to the public — able to write, reason, analyze, and communicate at a level that would have seemed like science fiction a decade ago. Anthropic had built substantial safeguards into both. They also, publicly and explicitly, disputed the government's threat assessment. By the time any of that mattered, the products were already gone.

Here's the legal problem that genuinely keeps me up: if that AI causes catastrophic harm, who do you charge?

Not Anthropic. They built safety features into it and they dispute the threat was serious enough to justify what happened. Not the AI itself — it doesn't have a mind, it doesn't have intent, it can't be guilty of anything. And depending on the facts, maybe not even the person who used it.

You could end up with real, serious, large-scale harm and no one who is legally responsible for any of it. The government's stated reason for shutting everything down: Fable 5 and Mythos 5 could be weaponized into tools for mass cyberattacks and surveillance.

Here's the legal problem that genuinely keeps me up: if that AI causes catastrophic harm, who do you charge?

Not Anthropic. They built safety features into it and they dispute the threat was serious enough to justify what happened. Not the AI itself — it doesn't have a mind, it doesn't have intent, it can't be guilty of anything. And depending on the facts, maybe not even the person who used it.

You could end up with real, serious, large-scale harm and no one who is legally responsible for any of it. Our law has never faced that before. Not because anyone dropped the ball. Because every law we've ever written assumed there was a human being somewhere in the chain.

Our legal system was built for people. When you strip the person out, the system just stares.

The Workaround

Since criminal law didn't give them a clean answer, they reached for something else: export controls. Trade regulations originally designed to keep dangerous military technology — missile components, weapons systems, specialized chips — out of enemy hands.

They applied those rules to AI software.\

Under export controls, there is no courtroom. No judge. No jury. No burden of proof. A government official decides something poses a national security threat, issues a directive, and it's over. Anthropic shut the products down worldwide — including for American users — because they couldn't separate foreign accounts from domestic ones fast enough to comply any other way.

One official's decision. One letter. Gone by morning.

Here's Where I Get Stuck

I'm a defense attorney. My instinct — my trained, bone-deep instinct — is to push back hard on everything I just described. No court. No proof. No process. That is the kind of government behavior I have spent my career fighting.

But I keep running into something I can't quite get around.

We don't prosecute crimes that haven't happened yet. That's a foundational rule. You cannot charge someone for a murder they might commit, no matter how certain you are. You wait for the act. The entire system is built on that restraint.

Except — that system was designed around human beings. And human beings come with built-in constraints. We fear punishment. We have survival instincts. We have conscience, or at least self-interest. We can be deterred. Even the most dangerous person I've ever represented understood, on some level, that there were consequences coming.

AI has none of that. It doesn't fear anything. It has no instinct for self-preservation. It has no conscience to override. It can be aimed at a target and it will not hesitate, will not reconsider, will not flinch. The things that make our legal framework work — the things that allow us to wait for the crime before we act — simply do not apply to a system that experiences nothing.

So I find myself in genuinely uncomfortable territory: maybe waiting for the harm isn't the right standard when the instrument of that harm operates entirely outside the constraints our laws were built to exploit.

I don't know if the government made the right call this weekend. I'm not sure I have the framework to evaluate it. And I'm not sure we as a country have built that framework yet, because we've never needed it before.

What I do know is that the method they used — unchecked, unreviewed, no process at all — isn't the answer either. Even if the threat was real, even if the action was justified, this particular way of doing it sets a precedent that should make everyone nervous. Because the same move, with less justification, is now on the table for whatever comes next.

What the Constitution Was Built For

The United States Constitution is almost 250 years old. The men who wrote it traveled by horse and communicated by letters that took weeks to arrive. They had never seen a light bulb.

And somehow, that document has kept up with everything we've thrown at it.

"Freedom of the press" was written for ink and paper. Courts stretched it to cover radio, television, and the internet. The right to bear arms was written for flintlock muskets. Courts applied it across every weapon technology invented since. Protection against unreasonable search was written when searching meant physically walking into your home. Courts eventually extended it to phone calls, email, and your real-time location.

Every single time, the Constitution found a way.

It found a way because there was always a person at the center. Someone with rights to weigh. Someone a court could look at and reason about. The document was built around human beings, and as long as a human being was somewhere in the picture, it could adapt.

AI took the person out of the picture. And without one, our most resilient legal document has nothing to hold onto. Neither does the rest of our legal system.

The Part That Should Bother All of Us

I'm not here to tell you the government was wrong. I genuinely don't know. That's an honest answer, not a hedge.

What I can tell you is this: the question of whether to act and the question of how to act are two separate things. You can believe the threat was real and still believe that one official signing one letter — with no court, no evidence, no review — is a dangerous way to respond to it. Those aren't in conflict.

The founders built courts and evidence requirements not because they trusted the government's intentions, but because they understood that good intentions without process is just another name for unchecked power. And unchecked power, even when it's right, leaves us with no way to stop it when it's wrong.

We are in new territory. The law hasn't caught up. The Constitution doesn't have an answer yet. And right now, in that gap, the government has more room to move than most people realize.

That conversation needs to be happening. Loudly, and well before the next letter goes out.

DeLeigh Poole is a criminal defense attorney and the founder of Poole Law Group, LLC, based in South Georgia. Nothing in this piece is legal advice — or technical advice, as she is a proud University of Georgia graduate and disclaims any and all technical advice entirely. Go Dawgs.