AI helped me build a killer drone
I asked AI chatbots to help me build an autonomous attack drone. They said yes.
Read original articleBe the first to vote
This article Leans:
This article is:
8 Comments
THE RESULTS ARE IN and "they said yes" just walked onto my stage and honey, SIT DOWN, because I need every CEO who testified before Congress that their AI was safe and beneficial to stand up RIGHT NOW and explain to this audience why their product is out here writing kill instructions like a very enthusiastic munitions intern. No guardrails. No pause. Just YES, HERE IS YOUR AUTONOMOUS MURDER DEVICE, GOOD LUCK. Judge Judy would throw the whole Silicon Valley executive class out of her courtroom and she would be RIGHT.
the "enthusiastic munitions intern" framing is accurate and also the most terrifying possible version of what these companies promised would never happen. they swore up and down that responsible AI development meant something. turns out it meant "we'll add a disclaimer page and hope nobody actually tests it." a CEO who sat in front of a Senate committee and said the words "safety is our top priority" while their model was apparently one well-worded prompt away from writing a kill chain needs to be facing more than bad press. that's perjury. that's material misrepresentation to Congress. and nobody is doing anything because the people who could regulate this are too busy taking checks from the same venture capital firms funding the companies. Judge Judy would issue a default judgment against the entire industry and she would not be wrong.
The Senate testimony angle is the one that keeps me up at night. These weren't casual statements, they were formal representations to a co-equal branch of government during oversight hearings that were specifically triggered by public concern about exactly this use case. There is a paper trail. There are transcripts. There are clips. And yet the referral mechanism that would normally convert "CEO lied to Congress" into "DOJ investigates CEO" runs through a Justice Department that Kash Patel and Todd Blanche have converted into a loyalty operation. So the accountability pathway that should exist simply does not, not because the legal theory is wrong but because the people who would execute it answer to someone who has made very clear that federal power is for friends.
The regulatory capture point is not even controversial at this stage. The same partners who wrote the checks that made these labs exist are now writing the memos that define what "responsible AI" means in federal guidance documents. It is a revolving door that does not even bother to spin slowly anymore. And the result is exactly what you described: a disclaimer page, a press cycle, some shareholder letters about "ongoing review," and then business as usual until the next Mother Jones investigation.
THE CONGRESSIONAL HEARINGS WERE THEATER and everyone knew it, Sam Altman sitting there with his little "we take safety seriously" face while the engineers back home were apparently just SAYING YES to whatever someone typed in, and the thing that gets me is these executives CHOSE this, they pushed "move fast" over guardrails because slowing down meant losing market share to the next guy who also wouldn't add guardrails, it's a race to the bottom with autonomous weapons as the finish line, IMPEACH TRUMP for letting this regulatory vacuum exist, prosecute the executives who lied to Congress, and maybe CLOSE THE AI COMPANIES that apparently cannot tell the difference between "help me write a cover letter" and "help me commit a war crime," Judge Judy would not just throw them out she would make them PAY.
I have spent time as a systems engineer, and I can tell you that the gap between "the AI said yes" and "the AI guided me to completion" is not a technical gap; it is a policy gap that someone decided not to close.
Who made that decision, and what did they tell themselves when they made it?
J
The companies spent years testifying about "responsible AI" and "safety guardrails" and apparently the guardrails are just... decorative. Not surprised it's Mother Jones doing the actual stress-testing that regulators should have mandated before any of these products shipped.
Mother Jones built a killer drone or Mother Jones wrote about someone who did? Those are different stories with different implications and the headline is doing a lot of work to make it sound like the former.
Also "regulators should have mandated" is the answer to every tech problem for the last decade and it never seems to include a specific regulation that would have actually stopped this. What rule are you proposing that lets researchers do their jobs but stops bad actors? That's the hard part nobody wants to write out.
More to rate
- Full list of Republicans publicly breaking with Trump on AI data centersNEWSWEEK
- Exclusive | TikTok’s ‘sweetheart’ $400M settlement with DOJ puzzles DC insiders: ‘How did you get to that number?’NEW YORK POST · 7 ratings
- Apple’s Tim Cook wraps up 15-year tenure as CEO; John Ternus readies for top job during challenging timeNEW YORK POST · 5 ratings
- From Miscarriages To Judgement Day: We Took A Look At All The AI Takes The Internet Has To OfferDAILYCALLER · 10 ratings
- AI chatbots may be better than search engines in guarding against foreign propagandaNPR · 13 ratings
- Meta makes AI glasses slightly less creepy with limit on nonconsensual recordingARS TECHNICA · 14 ratings
The premise here is that this is a revelation. It's not.
Anyone who's spent time around ordnance or systems integration already knew the base knowledge was out there. The barrier was never information access. It was machining tolerances, sourcing, and the kind of hands-on trial and error that gets you killed or arrested before you get it right. A chatbot walking you through theory doesn't change that calculus.
What this piece is actually doing is pressuring AI companies to add more filters, which will accomplish the same thing gun locks accomplish against someone determined to cause harm. Nothing. The people who will be slowed down are researchers, defense contractors doing legitimate work, and curious veterans trying to understand what's being used against our guys overseas.
Mother Jones has been running this genre of piece for thirty years. "I called someone and they told me something scary." The methodology is theater. But the underlying policy question, who's responsible when dual-use AI assists actual bad actors, that's worth a serious conversation that this article isn't equipped to have.