If your child’s school suddenly announces more in-class assessments and less weight on homework, it’s not random. Schools are grappling with a fundamental question: how do you evaluate learning when students have access to AI that can complete most assignments in seconds?
Andrej Karpathy, former director of AI at Tesla and a founding member of OpenAI, recently addressed a school board about AI’s implications for education. His framework offers parents a clear picture of how schools are likely to respond—and what it means for your family’s homework routine.
The Core Problem: You Cannot Detect AI Use in Homework
Karpathy’s first point is blunt: schools will never be able to reliably detect whether students used AI to complete homework. AI detection tools don’t work consistently, can be easily defeated, and are “in principle doomed to fail.”
This means schools must assume that any work done outside the classroom has potentially used AI assistance. The implication? Homework as we’ve known it—graded assignments completed at home that significantly impact final marks—may be fundamentally changing.
The Solution: Flip the Classroom Model
If schools can’t trust at-home assignments, the logical shift is toward in-class evaluation. Karpathy suggests the majority of grading should move to work completed in physical classrooms where teachers can monitor students directly.
This doesn’t mean homework disappears. It means homework might become practice rather than assessment—a place where students can use AI to explore concepts, get unstuck, or work through problems. But the real evaluation happens in class, without AI.
The motivation structure changes: students know they’ll be tested on their ability to solve problems independently, so they remain motivated to actually learn the material rather than just submit AI-generated work.
What This Looks Like in Practice
The philosophy mirrors how schools handled calculators: teach foundational skills first, then introduce the tool. Students learn arithmetic by hand so they understand what calculators do and can catch errors when they mistype.
AI requires the same approach, but the stakes are higher. Calculators rarely make mistakes. AI can confidently produce wrong answers, misinterpret instructions, or miss important nuances. Students who understand underlying concepts can spot these errors. Students who’ve only ever used AI cannot.
The Creative Middle Ground
Karpathy notes that teachers will have discretion in how they design assessments. The options include:
- No tools allowed (traditional closed-book exam)
- Cheat sheets permitted
- Open book
- Provided AI responses that students must evaluate or improve
- Direct internet or AI access for specific tasks
This creates flexibility. A literature class might provide an AI-generated essay and ask students to critique it. A maths class might allow AI for complex calculations but require students to show they understand the underlying logic. A history class might permit AI research tools but require in-class synthesis.
What This Means for Parents
These changes are emerging in schools globally, though implementation varies. US and UK schools with more autonomy may shift faster than centralised systems in Europe or Asia. Regardless of where you are, the underlying challenge remains: schools cannot verify at-home work anymore.
Homework is changing from assessment to practice. Your child may still have assignments, but they’re learning tools rather than high-stakes evaluation. Using AI to get unstuck might be acceptable—even encouraged. But the real measure of learning happens in class.
Expect more in-class assessment. If schools can’t verify at-home work, evaluation shifts to supervised settings. More quizzes, in-class essays, and monitored tests.
Ask your school about their AI policy. Schools are navigating this in real time. Questions worth asking:
- What’s your policy on AI use for homework?
- How are you adjusting assessment methods?
- Are you teaching students how to use AI effectively?
- How do you ensure students can work without AI when needed?
The Bigger Picture: Proficient With AI, Capable Without It
The goal isn’t to ban AI or pretend it doesn’t exist. It’s to ensure students can function in both scenarios: using AI effectively when available, and solving problems independently when it’s not. Given AI’s power—and its propensity for confident mistakes—foundational skills and verification abilities matter more than ever.
What to Watch For
Schools globally are wrestling with these questions right now. Some are moving quickly to in-class assessment models. Others are experimenting with AI-integrated assignments. Still others are trying to maintain traditional homework structures whilst acknowledging they can’t enforce them.
Your child’s school may not announce “we’re responding to AI” when they change assessment methods. But if you see shifts toward more in-class work, different homework expectations, or new policies about technology use, AI is likely the catalyst.
The parents who understand this shift—and the reasoning behind it—will be better positioned to support their children through the transition.
The Bottom Line
Schools are facing an unprecedented challenge: how do you assess learning when AI can complete most traditional homework? The emerging answer is in-class evaluation where teachers can directly observe student work.
For parents, this means adjusting expectations, understanding why assessment methods are shifting, and asking clear questions about your school’s AI policies. The goal: children who are proficient with AI whilst remaining capable without it.
