· via The Verge
Cody Wilson claims first AI-designed gun with WASP, a 3D-printed .22 handgun
Cody Wilson's Defense Distributed has shown off the WASP, a .22 caliber 3D-printed handgun it says was designed through dialogue with an AI model, lowering the technical barriers to unserialized ghost guns.

Cody Wilson, the activist behind the first 3D-printed firearm more than a decade ago, is now claiming another first: a gun designed by artificial intelligence. According to The Verge, which attended a demonstration at Defense Distributed's headquarters in Austin, Texas, the weapon is a single-shot .22 caliber 3D-printed handgun called the WASP, developed through plain-language conversation with a custom AI model and, like other ghost guns, carrying no serial number.
A proof of concept that fires live rounds
Physically, the WASP is rudimentary: The Verge describes it as essentially a pistol barrel attached to a spring-loaded plastic box, resembling Wilson's earlier single-shot Liberator but with an oddly rounded grip ill-suited to a human hand, a hallmark of early generative design. The design work came from a model Wilson calls GatGPT, trained on DEFCAD, the online repository of digital gun files he launched in 2012, which The Verge describes as the world's largest. Beyond designing, the model answers firearms questions and helps builders choose parts.
The entire process, Wilson claims, including the name and safety fixes after an early grip design risked costing someone a finger, resulted from natural-language dialogue between the model and its user. Wilson resists calling the WASP an AI-generated gun, arguing instead that what he has built is an AI capable of generating guns. Chat logs viewed by The Verge show the project began when an employee asked the model to create a new gun based on an existing design, the Ergo Harlot, framing the task as a challenge that required something fully functional with a unique design. The model responded with the WASP and submitted it for approval.
A demonstration with a misfire
At the test, the WASP was clamped to a do-it-yourself mount outside Defense Distributed's offices, with Wilson pulling the trigger remotely by a long string from behind a police riot shield. The first attempt produced only a loud click: Wilson said the striker had struck the cartridge without setting off the primer. After some troubleshooting, a second pull fired with a sharp crack about as loud as a firecracker, leaving a hole in a cardboard target downrange. Wilson then fired the gun by hand and missed high, making the AI-designed weapon two for three on shots successfully fired. Functionally suspect or not, The Verge notes the WASP fires real, lethal rounds, and Wilson expects such designs to improve as the model ingests more data and users interact with it.
Detection software meets a moving target
The safety concern is less this specific gun than the lowered barrier it represents. Producing a 3D-printed firearm has always demanded some technical know-how; now a novice could plausibly make one with a handful of prompts. Justin Wagner, senior vice president for law and policy at Everytown, told The Verge that Wilson's aim is to let anyone, regardless of background-check status or legal eligibility, design and manufacture weapons, which he called dangerous.
New York and California have recently passed laws requiring 3D printer manufacturers to install software that detects guns or gun parts and stops them before printing, though experts have questioned how or whether those methods would actually work. Wilson is already pushing back: in August he began releasing a tool called Hochulization, a jab at New York Governor Kathy Hochul, that subtly alters a file's underlying hash, a kind of digital fingerprint, defeating detection systems that rely purely on hashes. By late September, he claimed users had run files through the tool more than 26,000 times. More sophisticated detection uses AI to predict whether a file is likely a gun before it prints, but The Verge notes it is unclear whether such software, still in development, would reliably recognize a sufficiently strange-looking AI-generated design that exists in no database. A spokesperson for Hochul said people who intentionally thwart safety regulations put communities at risk, and that the governor remains committed to combating ghost guns and illegally manufactured firearms.
An old legal fight gains a new layer
Under US federal law, licensed gun makers must serialize the frames or receivers they sell. Wilson and others in the 3D2A movement argue that gun files are not actually guns but computer code, a form of protected speech. That argument has run through the courts for over a decade, often with Wilson as a party, and could yet reach the Supreme Court.
Why it matters
The WASP itself is a crude proof of concept that misfired on its first try. But it marks a shift: generative AI is no longer just helping people navigate existing gun files, it can propose novel, functional weapon designs on request. That collapses the technical skill barrier that has historically limited homemade firearms, and designs that never existed before undermine the detection approaches regulators in New York and California are betting on. The decade-old fight over whether gun files are speech now overlaps with a newer question about what AI systems should be permitted to help build.
- #ai
- #3d-printing
- #regulation
- #ghost-guns