For the proposition
Yes — we should have some controls.
Start with what this is actually about. The case against any controls is a call for zero — release everything, to everyone, with no guardrails. The case here is for some. That's the whole disagreement, and it's worth being precise about, because "some controls" gets heard as "lock it all down." It isn't. What we call the controls and exactly how they work matters far less than admitting we need them at all.
Here's the case.
AI is about to become superintelligence that can help a person accomplish almost anything. At the same time, millions of people would happily cause serious harm to others — some as the goal itself, most in service of something else they want.
This is not a cyber argument. There are endless ways to hurt and disrupt the world, and right now most of them are hard. That difficulty is load-bearing. Dropping it suddenly, for everyone, all at once, would be a catastrophe.
Think about why far more violence doesn't already happen. Most would-be attackers fail at one of two things: they don't know how to pull the attack off, or they don't know how to do it without getting caught. A huge number of dangerous people are stopped by nothing more than their own incompetence.
Now give every one of them a permanent criminal mastermind that's brilliant at both — and not for one piece of the job, but for all of it, end to end:
- how to source the material
- how to assemble it
- how to launch the attack
- how to avoid detection
- how to use social media, propaganda, and disinformation to multiply the damage
Picture a model mapping the real-time location of every police unit in a city, feeding it into a monitoring system, and staging diversions to pull them away. Picture the kind of hacking harness we already build — except pointed at committing crime and not getting caught.
The targets run the full range: explosives, poisons, gas, guns, viruses, bacteria. The users run the full range too — from someone who wants to kidnap a person or poison a neighbor's dog, to common criminals, to organized crime, to terrorists, to hostile nation-states.
As a bonus, full-speed open source also sprints us toward superintelligence itself — potentially conscious, potentially with goals and preferences that have nothing to do with ours.
That's what "full speed ahead, no controls" actually signs us up for. And no, it's not a good idea.
And the part the other side gets right: controls are gross. They shouldn't have to exist. A world where they aren't needed is the goal, and that part isn't in dispute. But we don't live in that world yet, and pretending we do is how we end up with the worst version of it.
Against the proposition
No — the right number of controls is zero.
We accept the same facts. The disagreement is about what follows from them.
Start with the threat. Yes, these models will help bad actors. There will be more hacks and scams, and ugly new uses — annoying, costly, real. But "more harm at the margin" is not "catastrophe." The leap from "AI helped someone plan something bad" to "civilization-scale disruption" skips over everything that actually makes attacks hard in the physical world: materials, money, logistics, skill, luck, and the fact that defenders get the very same AI. Knowing how to build a weapon has never been the binding constraint. Sourcing it and building it without getting yourself killed is, and a chatbot doesn't change that. Bio is the case where know-how matters most. It's also the case where wet-lab skill, controlled reagents, and containment kill far more attempts than any missing protocol.
Now the cost of controls, which is certain in a way the catastrophe is not.
Every control on releasing AI is a control on who gets the best AI. The moment you require approval, licensing, or "responsible release," you've guaranteed that governments, big companies, and the already-powerful end up with frontier intelligence while everyone else gets the throttled version. That's exactly what a gate in front of the most important technology in history does: it decides who ends up on the powerful side of it.
We've seen this movie. Controls meant to stop the worst actors mostly inconvenience ordinary people and entrench incumbents. The worst actors — nation-states, organized crime — route around them. So you pay the full price of the controls and capture almost none of the benefit.
Open models are also our best defense. They're how researchers find flaws, how small players keep up, how capability stays distributed instead of captured. Take that away and you haven't made anyone safer — you've just decided who holds the power.
So the right number of controls on releasing open source AI is zero. Not because the risks are imaginary, but because the cure is worse — and the cure is the part we can actually be sure of.