› Grades 9–12 · ages 14–18 · Policy debate
Who owns what the model learned?
“Training an AI model on copyrighted work without permission should require a license.”
A policy debate on training data and copyright, with the affirmative proposing a licensing plan and the negative attacking it or offering a counter-plan. Students meet fair use, the open-weights question, and the practical problem that the training already happened.
60 minutes · one class period · free to photocopy
Students will be able to
- Explain what training data is and why copyright law did not anticipate it.
- Propose or attack a concrete policy with attention to enforcement, cost, and who benefits.
- Evaluate the fair-use argument on its merits rather than on who is making it.
Vocabulary
- Training data — The examples a model learned from — often the whole internet, licensed corpora, or both.
- Copyright & training — Who owns the inputs models learned from — and who owns the outputs.
- Consent (data use) — Whether people agreed to their data being used to train or fine-tune systems.
- Open weights / open models — Models whose trained parameters are published for anyone to download and run.
- Closed / proprietary model — A model you access through an API — weights not public.
- Moat (competitive advantage) — Why one AI product keeps customers when models commoditize.
Agenda · 60 minutes
| Minutes | Segment | What happens |
|---|---|---|
| 0–6 | Warm-up | Human reader vs. model trainer. What changed? |
| 6–18 | Background + policy menu | Read the sheet. Teams map the four policy options on a grid: who pays, who benefits, who can enforce. |
| 18–28 | Case construction | Affirmative builds a plan (which policy, who administers, rate, exemptions). Negative prepares attacks and an optional counter-plan. |
| 28–52 | Policy debate | Aff constructive 3 · Neg cross-ex 2 · Neg constructive 3 · Aff cross-ex 2 · Neg rebuttal 2 · Aff rebuttal 2. Judges score plan, evidence, clash. |
| 52–60 | Written position | Exit ticket: which policy, why, and who loses under it. |
Affirmative — require a license
Creators' work has value that trained a commercial product; the law should require permission and payment, as it does for music and film.
- A model is not a reader. It's a product built from the work at industrial scale.
- Collective licensing already works for radio, streaming, and photocopying.
- Without a license, the incentive to create the next generation of training data disappears.
- Opt-out puts the burden on the person harmed, which is backwards.
Negative — training is fair use, or the plan fails
Learning patterns from public work is transformative and non-expressive; a licensing regime would entrench the richest firms and kill open models.
- The model doesn't store or reproduce the works; it learns statistics — closer to reading than copying.
- Only the largest companies can afford to license the internet. That's a moat, not justice.
- Open-weights and academic models can't pay; the plan hands AI to three firms.
- Counter-plan: an opt-out registry plus a compulsory license fund with a fixed rate — payment without gatekeeping.
Discussion questions
- 1.openingIs a model 'reading' or 'copying'? What test would tell you?
- 2.coreWho benefits under each policy option — and who is locked out?
- 3.coreIf a licensing rule makes open models impossible, is that an acceptable cost?
- 4.coreThe models already exist. Does your plan handle restitution, or only the future?
- 5.closingChoose one policy. Name who loses under it and why that's acceptable.
How policy debate runs
The proposition is a concrete policy ("Our school should…", "The law should…"). Affirmative proposes a plan and defends it; Negative attacks the plan and may offer a counter-plan. Constructive speeches (3 min), cross-examination (2 min), rebuttals (2 min). Judges score on plan, evidence, and clash.
In grades 9–12
Stakes are real: college, jobs, creative work, civil rights. Students can steelman, cross-examine, and write policy.
Preview the full pack
Cover, teacher guide, and every student handout. Print teacher pages once and student pages one per learner.
On a phone, open the PDF in a new tab — in-page previews are unreliable in mobile Safari.
Standards
- CCSS ELA SL.9-10.1 / SL.11-12.1
- CCSS ELA W.9-10.1 / W.11-12.1 (argument with counterclaims)
- C3 Framework D2.Civ.13 (public policy analysis)
- ISTE Students 2.c (intellectual property)
More for grades 9–12
Taught it? Tell us how it went: send classroom notes · AI concepts field guide · Try the Activities Pass free · Commission a full unit