The AI Moment Is Getting Violent

---

Sam Altman’s gate is still charred from a Molotov cocktail. A 20-year-old from Texas threw it there, and the authorities found his manifesto—pages advocating violence against AI executives. This wasn’t some random arson. The guy had a list of names.

That’s not a crime story. That’s a status update on where we actually are.

For a decade, I’ve watched Silicon Valley pitch AI as humanity’s next chapter. Transformative. Inevitable. Probably good, probably fine, trust us. The conversation has been almost entirely technical and entrepreneurial: Can we build it? How fast? Who wins the race?

Now the conversation is getting real in the worst possible way. And it’s not just the bombing.

Photo by Joel Santos / Pexels

When the Weapons Narrative Becomes Official

Here’s what shifted: AI as military infrastructure went from think-tank speculation to active national policy.

The global AI arms race isn’t a metaphor anymore. China, the U.S., Russia—they’re not having philosophical debates about superintelligence. They’re building weapons systems. The comparison to the nuclear age isn’t cute framing by journalists; it’s how actual defense officials are talking about it now. We’re watching the same acceleration that happened in the 1940s, except this time it’s happening in months, not years.

The difference between 2014 (when I first started covering this stuff) and now is stark. Back then, Elon Musk and a few academics were warning about AI risk. Reasonable people called them paranoid. Now? The U.S. military is allocating resources. China is publishing strategy papers. Russia’s talking about autonomous weapons. That’s not speculation. That’s procurement.

When governments start treating something as an existential competition, non-state actors notice. They take it seriously too—just not in the way Silicon Valley wants.

The Texas attack on Altman’s home wasn’t some isolated incident by a lone lunatic (though it might’ve been that, tactically). It was a data point in a larger pattern: people are starting to believe the stakes are actually as high as the hype suggests. And some of them are drawing violent conclusions.

Here’s the thing that keeps me up: the AI industry spent a decade insisting these systems were safe because they said so. We were supposed to trust the process. Trust the companies. Trust that competitive pressure wouldn’t override safety concerns because the incentives would naturally align.

Then you get attacked at your house, and those abstractions evaporate real fast.

Photo by nappy / Pexels

The Blowback Is Structural, Not Accidental

Meta’s been quietly scrubbing ads from Facebook after losing a landmark addiction lawsuit in California. The company—which literally revolutionized behavioral targeting, which built an empire on making your brain the product—is now defending its choices in court instead of bragging about them at conferences.

Meanwhile, Roblox, sitting on 144 million daily users and a business model that depends on keeping kids engaged, is scrambling to implement age checks after parents figured out the obvious: free-to-play gaming ecosystems optimized for engagement have a slight problem with child safety. The company’s now splitting into age-specific accounts. Translation: they finally admitted the old system was a mess.

These aren’t abstract regulation debates anymore. They’re not “oh, we need to think about governance eventually.” They’re real money, real lawsuits, real product changes. The blowback has teeth.

And it’s worth noting: these companies got ahead of it. Barely. Meta lost a trial before changing course. Roblox expanded age checks after controversy. Compare that to what happens if we get the AI trust model wrong, and you start to understand why people like the Texas guy might feel like the normal channels aren’t working.

I’m not justifying the violence. Let me be painfully clear. But I think it’s worth understanding the logic: the person throwing the bomb believes the institutions aren’t responding proportionally to the threat. Whether he’s right about the threat is a separate question. He’s definitely right that institutions move slowly.

The Geopolitics of Desperation

Here’s what worries me more than any individual attack: the arms-race logic is now irrevocable.

Once the U.S. and China both treat AI as strategic weapons infrastructure, you can’t un-ring that bell. You can’t have a reasonable conversation about slowing down, about safety research, about international agreements. Because the moment one side pauses, the other gets ahead. This is basic game theory, the kind that’s been driving military escalation since at least 1945.

The nuclear analogy holds up uncomfortably well. In the 1950s, after Oppenheimer started having second thoughts, the U.S. kept building bombs anyway. Not because anyone thought it was smart. Because the alternative seemed worse. The logic was: if we stop, they don’t, and then we lose.

That’s where we are with AI now. And unlike nuclear weapons, which are at least geographically distributed and require massive infrastructure, AI is distributed, reproducible, and getting cheaper to run every six months. The control problem is exponentially harder.

This is one of those moments where I genuinely don’t know what happens next. The incentives all point toward acceleration. The warnings are credible. The stakes are genuinely unprecedented. And the person you’d think would regulate it—governments—are all trying to win the race. The referee is competing.

The Stranger in Your Living Room

Elon Musk showing up on TikTok and Instagram is funny as a headline. It’s less funny as a symptom.

The guy owns X, he’s launching SpaceX soon, and he’s apparently testing social platforms just for fun. But what’s actually happening is the boundaries between platforms, between public figures and private citizens, between global communication infrastructure and personal brands—they’re all dissolving. Musk can use TikTok. TikTok can exist in America. There’s no coherent framework anymore.

That matters because it means there’s no obvious backstop. No institution saying “okay, we’re drawing a line here.” The AI companies are building god-level technology. The governments are racing to weaponize it. The public is starting to push back with bombs and lawsuits. And the mechanisms that usually slow things down—regulatory capture, institutional review, market friction—are either broken or actively accelerating the process.

Mark Zuckerberg’s AI sunglasses exist. They work. They’re weird and people hate them, but they work. In two years, they’ll be better. In five years, they’ll be ubiquitous. And someone will figure out how to weaponize the live facial recognition data flowing through 500 million of them. That’s not cynicism. That’s just how capability deployment works.

What I’m Watching

The next AI executive attack, and whether it’s a copycat or an organized signal — Altman’s house got hit, and there was a manifesto. If there’s another incident in the next 12 months—especially if it’s coordinated—that’s not a security issue. That’s the beginning of a movement. The Texas case goes to trial; watch whether the defendant’s manifesto gets public platform or gets legally suppressed. That’ll tell you what institutions think is contagious.

Whether Europe actually competes in quantum or just publishes good papers — The headline mentioned Europe could win the quantum race. They could. They won’t. Quantum computing is the one AI-adjacent technology where the U.S. and China haven’t completely colonized the field. If Europe fails here despite having resources and talent, it’s a tell that geopolitical competition has calcified into a two-player game. Watch for concrete quantum computing breakthroughs from European companies by Q4 2025. If they don’t exist, the race is effectively over.

Meta’s legal defense strategy in the next addiction case — They lost one trial. There will be more. The interesting question isn’t whether they lose again (they probably will). It’s whether they argue that social media addiction is a feature of engagement-driven systems (which would implicate the entire industry) or that Meta specifically failed to implement safeguards (which is survivable). Their legal strategy will tell you whether they believe regulation is coming or whether they think they can litigate their way through it.

How many months until an AI system gets blamed for a military incident — Not used in one. Blamed for one. A drone kills someone, the military says their AI system made a targeting decision it wasn’t supposed to make. That’s the moment the geopolitical logic breaks. Everyone suddenly wants regulation again, but by then the systems are too integrated to pause. Watch for this in the next 18-24 months of Middle East operations or Ukraine. When it happens, that’s when the real conversation about control starts.

The bomb at Altman’s house is a warning label on the product. The question is whether anyone reads the label before opening the box.