AI takeover may look less like Skynet and more like people slowly unlearning how to think. Why attention is the defense.
September 24, 2026

Originally posted by Kenneth Schlenker on LinkedIn.
Thought experiment: if we (humans) manage to stay in control of AI, who will still be smart enough to do it?
I use various AI models daily, and people around me use it for coding, therapy, shopping, relationship advice, grading and more. Can we still think independently, without it?
I found this study interesting: researchers asked 1,339 teachers to evaluate student work that had been paired with an intentionally incorrect grade. Half were told the grade came from another teacher; the others were told it came from AI. When the grade was unfairly harsh, teachers who believed it came from AI corrected it less aggressively, leaving behind an error that was 22 percent larger. AI sounds authoritative, so it's hard to challenge.
I don't think "AI takes over" looks like Skynet (with terminator AI robots killing people). It's probably going to be a lot more boring and banal: more and more people unlearn how to think for themselves, accept what AI tells them because it sounds smart, and don't have the skepticism, confidence, patience, or sustained attention to challenge it.
AI doesn't need to seize control from humans because we are gradually surrendering to it and we're not aware of it.
So when I think about human oversight of AI, I don't see a "smart person in charge" somewhere at the UN or a world organization. I think human oversight needs to come from a population that can actually notice when something doesn't add up, and feels confident enough in its own judgment to push back.
This is a skill, it has to be learned and practiced. Maintaining our attention is key to it.
This is one of the reasons I feel so strongly about what we're building at Opal over the next ten years.
We're here to protect people's attention and their sense of agency.
