Articles & Essays

California's 3D Printer Law Passed. The Part Everyone Fought About Didn't.

AB 2047 asks whether a 3D printer can be built to refuse a gun - and gives a standards body until 2029 to answer. For anyone who runs printers for a living, that question is the whole story.

On September 26, Governor Newsom signed AB 2047, the California Firearm Printing Prevention Act, written by Assemblymember Rebecca Bauer-Kahan. It was one of ten gun bills he signed that day.

The goal is to make 3D printers sold in California eventually carry "firearm blocking technology": hardware, firmware or other built-in measures that run a print file through a detection algorithm and stop the job if the file is a gun or a machine-gun conversion device (3DPrint.com).

Nothing changes for anyone today. Here's how the law actually moves:

Starting July 1, 2027, the state Department of Justice has to check every three months whether ASTM International - the standards body - has published a standard for this kind of technology. If one exists and DOJ decides it's feasible, DOJ gets 24 months to write regulations. The sales restriction starts one year after that.

And if ASTM hasn't published a qualifying standard by July 1, 2029, DOJ's duty to keep checking simply ends.

What came out of the bill on its way to the governor matters just as much. Earlier versions had a per-model attestation form, a public list of approved printers, civil penalties up to $25,000 per violation, and an exemption for industrial users in aerospace, medical and automotive work. All of that is gone. What's left is an exemption for entertainment-industry props, and room for DOJ to add others later.

That is a much narrower law than the one people were fighting about in the spring. Credit to the people who pushed back - and credit to the authors for listening. New York passed something similar earlier this year, and it likewise requires nothing until a working group studies whether it can be done.

Here's why it matters anyway. As far as I can find, this is the first state law that puts the printer itself in charge of deciding what it's allowed to print.

And the limits, stated plainly: the core requirement is not active. It depends on a standard that doesn't exist yet and a feasibility call that hasn't been made. Anyone telling you "California now requires print-blocking" is wrong. It doesn't, and it may never.

There are two loud takes, and they don't talk to each other.

One: California is stopping ghost guns at the source. That's close to the author's own framing, and the problem it points at is real. A printed gun has no serial number and no sale record, and police have been finding more of them.

Two: California wants your printer to spy on you. That one wasn't paranoid either. Earlier drafts, with attestation forms and a state list of approved machines, really did read like a surveillance mandate for a hobby that millions of people do in their garages.

Both sides have a point. And both are arguing about a version of the law that didn't pass.

What was signed isn't a mandate. It's a question, sent to a standards body, with an expiration date on it: can this be built?

That's actually a reasonable shape for a law. Ask whether something works before you require it. But it means the whole fight now lives inside that one question, and almost nobody arguing about it runs printers for a living.

So here's what "blocking technology" means in practice. Every file gets screened before it runs. That can only happen two ways. Either a classifier lives in the printer's firmware and makes the call locally, or the printer sends the file somewhere and waits for permission.

Both have a cost. A gun, broken into parts, is tubes, springs, brackets, grips and housings. So is a normal week of print jobs. A classifier looking at a shape with no idea who's printing it, or why, is going to get some of those wrong - and every wrong answer is a stopped job for someone who did nothing wrong.

I'm generally on the side of machine learning taking on more of our work. But the test for any model is what a wrong answer costs, and who pays for it. Here, the people paying would be the shops and hobbyists who bought a compliant printer and left the firmware alone. The person determined to print a gun is the one least likely to do either.

And the server version is worse. A printer that has to ask a server before it prints is a login screen bolted onto a machine you own. We just watched that go wrong for a video game in the same week. On a production line, "the server's down" means nothing ships.

If you print anything - a farm, an Etsy shop, a repair business, a garage hobby - this law isn't asking you to do anything yet. It's asking a question about you.

The answer won't come from Sacramento. It'll come from whoever helps write the ASTM standard, and from whatever DOJ decides counts as "feasible." Standards get written by the people in the room. If the room is only legislators and firearm-detection vendors, the standard will be built around what's easy to detect, not around what it costs to get it wrong.

And this goes beyond printers. Any time a law asks "can this be built?" instead of "build this," the real decision moves to whoever answers. That's usually a quieter room, with fewer people watching, and a much longer runway than the headlines suggest. This one runs until 2029.

A printer that has to ask permission before it prints is a different machine. Maybe there's a version of that worth building. Nobody knows yet - that's literally what the law is asking.

But if California wants an answer, the people who'd be asking that permission every day should be the ones helping to write it.