Class 7 · CBSE AI · Strand C — NLP, Vision, and LLMs Deep-Dive
AI, training data and copyright — the big debate for Class 7
AI learns from human-made work — but is that fair use? The lawsuits and ethics, for Class 7.
What this concept actually says
- AI training requires massive datasets of human-created work — the legal and ethical status of using this work without permission is actively contested
- The 'fair use' argument (training is transformative) is challenged by artists, writers, and coders whose work was used without consent or compensation
- Data governance — who owns training data, who benefits from it, and who can audit it — is a defining political question of AI development
An analogy your child will recognise
A chef who studied in others' restaurants
A chef who apprenticed in five great Indian restaurants, learning every technique and secret recipe, then opens a competing restaurant is doing something humans accept as normal learning. But if they secretly recorded every recipe and sold the recordings, that would be wrong. AI training sits somewhere uncomfortable between these — learning at massive scale, but from data that wasn't fully offered for that purpose.
A student who memorises the entire school library
A student who reads every textbook, novel, and newspaper in the school library is admired. But if they then start a paid tutoring business by reciting exact passages from those books, the authors might object — even if each recitation is slightly paraphrased. The question of when 'learning from' becomes 'profiting from without permission' is exactly what the AI copyright debate is trying to settle.
Common misconceptions to watch for
- If something is publicly available on the internet, it is legal and ethical to use for AI training — terms of service, copyright licences, and ethical norms often restrict such use even when content is technically accessible.
- Copyright only applies to exact copies — AI systems that reproduce style, tone, or closely paraphrased content raise genuine legal challenges even without verbatim reproduction.
Key facts in one breath
- The New York Times filed a lawsuit against OpenAI and Microsoft in 2023, alleging that GPT models memorised and can reproduce significant portions of NYT articles.
- The EU AI Act (2024) includes provisions requiring AI companies to disclose summaries of copyrighted training data used in general-purpose AI models.
- The concept of 'data poisoning' has emerged as artist resistance — intentionally modified artworks that, when used as training data, degrade the quality of generated images (e.g., the Nightshade tool).
- India's Copyright Act (1957) is currently being reviewed for amendments to address AI-generated content and training data — the legal landscape is not yet settled.
How Dhee Learning teaches this — the 3-stage question loop
Every Dhee Learning session for this concept follows three stages. We share the questions Dhee actually asks, so you can hear what a session sounds like.
Stage 1 — Surface
Imagine a new musician who listened to thousands of hours of AR Rahman's music without paying for it, learned Rahman's style completely, and now sells albums that sound just like him — but aren't copies of any single song. Has the musician done something wrong? Why is this question harder than it first sounds?
Rote answer
"Copyright is about not copying someone's work."
Understood
"It's hard because style isn't copyrightable, and all artists are influenced by others — but this musician had an unfair advantage by consuming the work at massive scale without any payment or acknowledgement. AI training does this at billions of examples of scale. The question isn't just 'is it legal?' but 'is it fair?'"
Stage 2 — Reasoning
GitHub Copilot is an AI coding assistant trained on billions of lines of public code, some of which is licensed under terms requiring attribution if reused. Explain the conflict between the 'scale makes it transformative' argument and the 'consent and attribution' argument from a developer's perspective.
Follow-up Dhee may use: Should AI companies be required to disclose what data they train on? What would the consequences of that requirement be — for them, and for the people whose data was used?
Stage 3 — Application
You're designing a new image generation AI for a school art competition platform. Propose a data sourcing approach that addresses consent, attribution, and fairness — and explain one trade-off your approach introduces.
Misconception Dhee watches for: Believing that publicly available = freely usable for training — public availability online does not imply consent for all uses; licences, terms of service, and ethical norms constrain what 'public' means in practice.
Related concepts
Want your child to actually understand this?
Dhee turns this concept into a 15-minute spoken session — asking, listening, and probing — so your child builds the idea themselves.
Frequently asked questions
What is the copyright and training data conversation — explained for kids? +
AI learns from human-made work — but is that fair use? The lawsuits and ethics, for Class 7.
What's the most common mistake children make about this concept? +
If something is publicly available on the internet, it is legal and ethical to use for AI training — terms of service, copyright licences, and ethical norms often restrict such use even when content is technically accessible.
How does Dhee Learning teach this in a Class 7 session? +
Dhee opens with a question — for example: "Imagine a new musician who listened to thousands of hours of AR Rahman's music without paying for it, learned Rahman's style completely, and now sells albums that sound just like him — but aren't copies of any single song. Has the musician done something wrong? Why is this question harder than it first sounds?" — listens to your child's answer, then probes the reasoning behind it. The session ends when the child can apply the idea to a brand-new situation, not just recall it.