In San Antonio, Texas, 22-year-old Nathaniel Michael Carrasco lived close enough to Serna Elementary School that he could probably see the playground from his window. According to the arrest affidavit, that proximity gave him ideas. Police say he spent part of August 11th typing out his plans to an AI chatbot instead of, say, a therapist, a friend, or literally anyone who might have called someone.
He’s now sitting in the Bexar County jail on a quarter-million-dollar bond, facing a third-degree felony charge for making a terroristic threat, and North East ISD is left trying to explain to a school full of worried parents why nobody told them sooner.
Carrasco is said to have told the chatbot he was “getting a gun as soon as I can,” and that once Serna Elementary started up for the year, he planned to walk in and start shooting kids. He asked it to search for images of dead schoolchildren. He also wrote that there would be “no escape, no matter what,” and made a point of saying he didn’t want the bot talking him out of any of it.
Strip away the fact that his audience was a chatbot instead of an anonymous forum post, and this reads like a form letter. It’s the same script, the same aesthetic obsession with being unstoppable, and the same cribbed language you’d expect from someone who’s spent a little too much time wallowing in True Crime Community content and imagining himself as the next name people whisper about.
Nothing in what he wrote suggests an original thought. It suggests a guy who found a script online and decided to perform it for an audience that happened to be artificial.
However, none of the coverage that I’ve seen explains how these chats actually got from that conversation into the hands of investigators. We know the FBI’s National Threat Operations Section looped in the Southwest Texas Fusion Center on August 28th, more than two weeks after the messages were sent, and we know SAPD then used his phone provider, his internet service, and his Google account to figure out who he was. What we don’t know is the step before all of that. Did Anthropic flag the conversation and report it? Did someone else stumble across it? Nobody’s saying. That’s a genuinely interesting gap in the story, and it’s one I’d want answered before drawing any conclusions about how well or poorly any of this actually worked.
What I can say, based on every other version of this story I’ve seen so far, is that there’s no indication the chatbot helped him with a single thing. Nothing in the reporting suggests it gave him tactical advice, pointed him toward a gun store, or offered any kind of assistance whatsoever. If history is any indicator, Claude more than likely refused to engage the moment things turned toward violence.
And yet, if you read the headlines on this case, including this one, you’d think the chatbot was the one facing charges. “Man used AI to plan mass shooting,” they say, like the software was an accomplice instead of a receptacle he happened to dump his intentions into.
I get why that framing sells. It’s new, it’s unsettling, and it lets everyone avoid the far less exciting and far more uncomfortable question of how a 22-year-old living alone next to an elementary school got this far into a violent fantasy without a single person around him noticing or intervening.
AI makes for a cleaner villain. It doesn’t have grieving relatives who’ll go on the news defending it, and it can’t tell its side of the story. But there’s a version of caution about artificial intelligence that’s actually warranted, and there’s this, and they’re not the same thing.
You can worry about AI taking your job or about the surveillance implications or the deepfakes and election misinformation and whatever fresh nightmare Silicon Valley cooks up next. Don’t worry about this particular chatbot being the reason a man wanted to shoot up a school, because that notion came first.
I’ve said this before, and I’ll probably say it again before the year’s out. Every generation finds its scapegoat technology, decides that this is the one that’s finally corrupting the youth beyond repair, and moves on to the next one once the panic burns itself out.
It was music, then movies, then video games, then the internet, then social media, and now it’s chatbots. The through-line in every single one of these moral panics is that the technology gets blamed for problems that existed long before anyone figured out how to code them.
Isolation, untreated mental illness, an obsession with infamy, and easy access to firearms do not require a chatbot to exist. It just requires a person willing to act on it and a society that keeps failing to notice until after the arrest.
Carrasco is the one who typed those words and decided a school full of little kids was an acceptable target for whatever he was working through. No chatbot made that decision for him, and no amount of clever prompt engineering changes where the responsibility ultimately lands.
If it hadn’t been an AI program, it would’ve been something else. A video game, a movie, a song, a website, or a forum. We’ll always find some object to point at that isn’t the actual gun, because pointing at the gun would mean having a conversation this country still refuses to have.
Until then, we’ll keep having this same argument, with a different piece of technology playing the villain and the same guns sitting untouched in the background, waiting for the next guy with a grievance and WiFi.
(Sources)
Leave a Reply