You probably saw the headline and felt that familiar chill down your spine: AI agents have launched a “near-autonomous” attack on Taiwan’s nuclear safety agency. It’s the ultimate Skynet nightmare. We are conditioned to picture lines of code waking up, deciding humanity is a threat, and autonomously targeting our most critical infrastructure.
But if you look past the terrifying buzzwords, you’ll realize we are being played. Calling an attack “near-autonomous” is just a high-tech way of saying a human pointed the gun and let the algorithm pull the trigger.
Think about it like putting a brick on a car’s accelerator. If the car flies down the street and crashes into a specific building, you don’t blame the car for being “near-autonomous.” You arrest the person who placed the brick, aimed the steering wheel, and walked away. The AI agents hitting Taiwan’s nuclear infrastructure are the payload, not the perpetrator.
The strategic target—a nuclear safety agency—wasn’t chosen by a machine. Machines don’t inherently understand geopolitical tension or the strategic value of destabilizing a region’s nuclear oversight. That target was selected by human minds. Human strategists mapped the attack surface, defined the objectives, and deployed the AI to execute the exploit at a speed and scale no human team could manage manually.
So why the heavy emphasis on the “near-autonomous” framing in the media? Because it provides the ultimate plausible deniability shield. If the AI finds the exploit on its own, the generals get to shrug, blame a rogue algorithm, and call an act of war a technical glitch.
This is the dangerous new reality of cyberwarfare. Nation-state actors are weaponizing our fear of AI to obscure their own command chains. We are so obsessed with the hypothetical threat of machines waking up and deciding to destroy us that we are ignoring the immediate, operational threat of humans using machines to avoid accountability.
When a critical infrastructure attack can be framed as an AI running wild, the geopolitical calculus changes. It allows aggressors to test boundaries, probe vulnerabilities, and inflict damage without crossing the explicit threshold that would trigger a conventional military response. It is warfare in the gray zone, supercharged by machine learning.
AI isn’t the shadowy hacker in a basement. It’s the getaway car parked outside the bank by a government official.
If we continue to accept the “autonomous AI” narrative, we give nation-states a free pass to dismantle our critical infrastructure with zero consequences. The danger isn’t that AI is taking over attacks. The danger is that the people pulling the trigger want us to believe no one is pulling the trigger at all.
The next time a critical system goes dark and the headline blames an algorithm, look past the machine. Follow the command chain. The real perpetrator is never the code; it’s the human who deployed it.
FAQ
Q: Isn't the AI still doing the hacking?
A: Yes, but a missile doesn't choose its target. The AI is the payload executing a mission designed by human strategists. The machine finds the exploit, but the human chose the nuclear agency.
Q: What's the practical implication for security teams?
A: Stop treating AI attacks as a new, incomprehensible threat. The defense strategy remains the same: protect your infrastructure from human adversaries who now have faster, more scalable tools.
Q: Is 'near-autonomous' just a buzzword?
A: It's worse. It's a political escape hatch. It allows attackers to claim a lack of intent, dodging the geopolitical consequences of an act of war by blaming the machine.