Evan Liu: AI's Real Risk Is Humans Surrendering Their Cognition

The AI risk debate rarely focuses on what happens inside people's heads. In a piece for Tech Policy Press, Harvard graduate Evan Liu argues that the real catastrophe is not autonomous machines turning on their creators, but what he calls "cognitive surrender": the tendency to adopt AI-generated outputs without a second glance, bypassing the friction of reasoning in favor of outsourcing the work of deliberation.

Liu's examples are concrete and recognizable. The student who once wrestled with a difficult text now asks Claude for a summary. The programmer who once debugged line by line now supervises generated code they barely understand. Over time, Liu writes, "the muscle of sustained thought weakens through disuse." The concern, he argues, is not that AI makes people less intelligent overnight, but that "constant reliance on artificial cognition changes our relationship to effort itself."

The piece ties this to the economic pessimism of Gen Z. According to Northwestern Mutual's 2026 Planning & Progress Study, almost a third of Gen Z are risking money on sports betting, prediction markets, and crypto; of those, eight in ten believe these assets offer a faster path to their goals than traditional methods. Liu's reading is that a generation that grew up through a global pandemic, a student debt crisis, and a climate emergency has rational cause to apply a heavy discount to any future payoff that requires decades of patient accumulation to realize. If conventional mastery now seems like a poor investment, AI as a cognitive shortcut can start to look like the only rational play.

What the essay does not give you is evidence that this atrophy is happening at population scale, or whether it is reversible. Liu makes a case from behavioral signals and logic, not longitudinal data. The question of whether AI-assisted shortcuts produce worse thinkers over time, or simply different ones, remains genuinely open.

The closing grounds things in Ecclesiastes: meaning-making through effort and struggle may be one of the last things AI cannot provide. Whether that is a reason for urgency or a source of reassurance depends on how seriously you take the idea that productive struggle is itself the point.