Linus Torvalds Finally Speaks: Should AI Write the Linux Kernel? [And a War Just Began]

[00:00:00]Six words from Lionus Stoaltz just set the entire Linux world on fire. He said Linux is not one of those anti- AI projects. Some are calling it the moment the Linux kernel opened its doors to AI.

[00:00:12]Others are calling it the start of an open-source civil war. But here's the thing, almost everyone got what he said wrong. And underneath that one sentence is a question the community has been fighting over long before towards ever weigh in. Should AI be writing the Linux kernel? Stick around because the answer to that is messier than the headlines suggest. So, let's set the record straight because what Lionus actually said and what the internet turned it into are two very different things. And the real story is bigger, not smaller.

[00:00:42]First, where it happened. This didn't come from a stage or a press release. It came from a public kernel mailing list.

[00:00:48]The exact place where the Linux kernel is actually built decision by decision every single day. And that's precisely why it carries so much weight. When Towards speaks here in the engine room of the kernel, it is scripture. And here's what lit the fuse. It started with an AI tool called Sashiko, an experimental system that uses AI to help review kernel patches. One contributor wanted nothing to do with generative AI.

[00:01:12]Others wanted to use the tool. Anyway, that single clash forced open a question the whole project had been circling for years. And that's the moment Lionus stepped in. And his point was really two things. One, Linux is not going to brand itself an anti-AI project. Two, no single person gets to veto the tools everyone else wants to use, but and this matters, he also made it clear that using AI stays completely optional and only if it actually reduces the workload on maintenance, not piles more onto them. Now, here's a line everybody skipped right over. Allow does not mean trusted. Even where AI is welcome, a human still has to understand the code, test it, and put their name on it.

[00:01:53]Nothing gets a free ride into the kernel just because a machine made it. So no, Linus didn't fling the doors open to botridden code. But he outright refused to make Linux anti- AAI as well while keeping every real responsibility right where it's always been on a human. But settling what Lionus meant doesn't settle anything because it drags the real question out into the open. The one the entire open source world is now being forced to answer. So here it is out in the open. Should AI generated code be allowed into the Linux kernel?

[00:02:25]Simple question, but the answer not so simple. And before you brush this off as some nerdy inside baseball debate, stop and think about what the kernel actually is. This isn't a hobby project on a dusty old laptop. Linux runs the servers behind the entire internet. It runs billions of Android phones, your car, your bank, the world's supercomputers, all of it running on this one kernel. So the question of what gets to go inside it and who we trust to write it isn't academic. It's about as high stakes as software gets. And there are two very real sides to this. On one hand, the maintenance are drowning. They are overworked. They are underpaid. And AI can genuinely help catching real bugs and writing working patches faster than any human ever could. On the other hand, that very same AI can turn out code that looks absolutely perfect and is absolutely confidently wrong at a scale no human team ever could hope to keep up with. So, should AI write the Linux kernel? Here's my honest answer. I am not going to answer that. Not yet.

[00:03:24]Because honestly, it's not really my call to make. But before we go any further, I have to be honest with you about something. Here's the truth. I was an AI skeptic, too. I genuinely thought this whole thing was just a glorified autocorrect. Then chat GPT started improving rapidly and slowly it won me over. Today is one of the most valuable tools I use. It gets me more done and it makes my life easier. But honestly, I miss the old way. My research, my creativity, my own mind doing the work.

[00:03:52]That meant something. I used to go on long walks thinking about things. I do miss the way I work before AI came along. While I still do a lot of the stuff on my own, AI helps. Here's the hard part. If I clung to that today, the competition would eat me alive. So, I adapted. Adapt or get left behind. And that same tug of war, the Linux community has been leaving it for years now. So, let's get one thing straight.

[00:04:16]That email from to wars didn't start this war. It just switched the lights on. The Linux community has been quietly fighting over AI for years now. And to understand where this is going, you have to understand where it's been. Let's start with Towalts himself because a lot of people are acting like July was some big conversion moment. It wasn't. Lionus has been remarkably consistent this whole time. This is the same guy who called the AI industry 90% marketing.

[00:04:42]The same guy who shrugged at Y coding.

[00:04:44]Sure, use it for your little toy projects, whatever. The same guy who openly mocked developers bragging about their AI generated code, but and here's the consistency. He's always back tools that genuinely make the work easier. So his statement wasn't a flip. It was the same principle he's held forever. Tools are tools. Humans stay accountable. Now the anti- AI camp and the biggest grievance isn't some philosophical hand ringing. It's that they are drowning.

[00:05:10]See AI can generate a patch or security report that looks completely legitimate way faster than a human volunteer can actually check it. Developers call this the review asymmetry problem. The machine writes in seconds what takes a human hours to verify. The poster child for this curl or C URL if you don't know it curl is a tiny tool that quietly runs on billions of devices. Its maintainer got so buried in fake imaginated security reports death by a thousand slobs he called it that he straight up killed the project's bug bounty program.

[00:05:41]A flood of confident plausible looking AI garbage overwhelmed the humans so badly they had to shut the whole thing down. For the developers watching this is the real cost. Verification is the expensive part and AI makes producing noise almost free and this is where the community splits right down the middle because there's no single rule here. Gen 2 has banned AI generated contributions outright full stop. Feder on the other hand allows responsible use. KO and Net PhD are barking over prominence. That scary little question of who or what actually wrote this code and can we even legally use it? And the software freedom conservancy has stepped in to defend a developer's right to refuse AI entirely without being pushed out for it. One ecosystem completely opposite laws. But and I have to be fair here, it's not all horror stories because sometimes the thing just works. Greg Crower Hartman, essentially Towels's right-hand man, has shipped AI assisted audit patches straight into the kernel, openly disclosed, human check, and genuinely useful. Mosilla and Antropic used AI to hunt down real bugs in Firefox that humans had missed. So under proper human validation, AI is catching things we couldn't. The skeptics aren't wrong, but neither are the people adopting it. So how does the kernel itself thread this needle? Simple. AI assisted code is allowed, but only through the exact same gauntlet as everything else. compatible licensing, human review, testing, and a human signature. The signed off by line certifying the developer certificate of origin. And here's the kicker, an AI cannot sign that. A human has to put their name on it and take the responsibility. This is allowed does not mean trusted in action. So here we are, both sides holding real legitimate evidence. Ban spreading on one side, adoption spreading on the other, positions hardening day by day, which leaves us with one very uncomfortable question. Is this heading somewhere much, much darker?

[00:07:31]So, is an open-source civil war coming?

[00:07:33]Here's my honest answer. No, I don't think so. There's no Linux folk, no mass resignations, and nobody's being forced to use AI against their will. But don't relax just yet, because this isn't peace either. What we've got is low, simmering tension, one that flares up with every new incident, and every flare up burns just a little hotter than the last. The word war is early. It isn't wrong. And here's the part everyone misses. This was never really humans versus machines.

[00:07:59]The real fight is about power. Who controls the tools? Who has to put their name on the work? And who pays the price of verifying whatever the machine spits out? That fight, it's already here. And the codes haven't even caught up.

[00:08:12]Legally, it's still a mess. Who even owns AI generated code? The ground everyone is fighting on is still moving.

[00:08:18]And in the open source world, this isn't some minor footnote. It's the entire foundation. The whole thing runs on license like the GPL. So if you can't prove who wrote a piece of code or where it came from or has ambiguity regarding the ownership or license, you can't legally share or build on it or trust it. And the entire chain that Linux is built on quietly starts to fall apart.

[00:08:38]Which brings me right back to where I started. So here's where I land. That tug of war happening in the community, it's really just my own little confession blown up to the planet's guild. Missing the old way is valid, but it was never a strategy. AI isn't going anywhere and Linux isn't surrendering to it either. It's doing what it has always done best, adapting on its own terms with a human still holding the pen.

[00:09:01]Adapt or get left behind. So now I'll throw it to you. Would you trust AI written code in the kernel? Something that the entire world runs on. Let me know down in the comments. All right. If you enjoyed this video, if you found it useful, definitely consider subscribing to the channel and also leave me a big thumbs up. By the way, if you haven't already, check out my course Linux Mastery Express. I've designed this course to level up your Linux skills very quickly. With this course, you'll get so comfortable using the terminal commands that your friends will think you're a Linux wizard. You'll get perfect with the most used, most useful commands and also learn advanced things like using the V editor and shell scripting as well. Linux Mastery Express link in the description. Do check it out. All right, next up, check out the top 10 hottest apps that you should be using in 2026. I got some really cool ones there, so definitely don't miss that. All right, this is Linux Tech signing out.