Most parents have already watched a child paste a homework question into an AI and copy the answer straight back out. The harder question is whether that child could tell when the answer is wrong. This month, the National University of Singapore answered it at institutional scale.
Starting in academic year 2026/27, every first-year undergraduate at NUS must complete a module called THE1008, Applied Generative AI: From Prompting to Evaluation. It runs in the first two weeks of term, before regular coursework begins. All students, faculty and staff also get access to ChatGPT Edu from 31 August 2026, piloted in selected courses first. NUS describes the goal for graduates as AI fluency, built through the common curriculum rather than parked in one optional elective.
The module title is the interesting part. Most AI education for children currently stops at the first half of it. Write a better prompt, get a better output, move on. Evaluation is the other half, and that is where the thinking actually happens. Can the student say what the model got right, where it is likely to be wrong, and how they would go and check? A university placing that in week one of a degree is saying it considers the skill foundational rather than advanced.
For a child between 9 and 15, this is easier to see than it sounds. An AI will happily produce a paragraph of history containing a confident date that never existed. A student with fluency notices that the date is carrying the whole paragraph and goes looking for a second source. Teens in LPA’s AI Game Studio program hit a sharper version of this: the model writes narration that contradicts what is actually happening in their game, and nobody else is going to catch it. The student has to spot the failure and constrain the system so it stops happening.
Parents can start this at home with one question, asked every time an AI is used: how would you check that? It costs nothing and it moves a child from consuming an answer to holding it up to the light.
LPA programs are built around that sequence. Students use real tools to build something real, then have to explain what the system produced and show where it could break. Universities are now making that a graduation requirement. There is no particular reason to wait until then to begin.