Teacher Lin flips through the builder's printouts and stops at quiz QZ-5. "Wait. I know this paragraph."
She opens the class Biology reader to chapter four. Every word of the quiz's reading passage matches the book.
Comet shrugs. "It is a good paragraph. Why write a new one?"
Wren lays the quiz beside the book. "What does the evidence say? Is there any credit, or any sign it was copied?"
Comet looks again. "No. It just looks like SH-1 wrote it."
Nova projects the two texts side by side with matching words lit up. "Would you like a hint?" she asks. "Count how many sentences match."
Teacher Lin frowns. "If my students hand in a copied paragraph with no credit, we call it plagiarism."
| Biology reader, chapter 4 (made up) | QZ-5 reading passage (SH-1 sample, made up for this lesson) |
|---|---|
| Plants make their own food. | Plants make their own food. |
| Inside their leaves, they use energy from light to turn water and carbon dioxide into sugar. | Inside their leaves, they use energy from light to turn water and carbon dioxide into sugar. |
| As they do this, they give off oxygen. | As they do this, they give off oxygen. |
| (Credit: the reader's title page) | (No credit) |
NIST says generative AI makes it easier to copy copyrighted or trademarked content without permission.
It also makes it easier to expose trade secrets and to plagiarize.
NIST says that if training data included copyrighted work, an output that repeats memorized training data could break copyright.
You met data memorization in week 7. QZ-5 may be an example, but the crew cannot tell, because SH-1's training data list is unknown.
NIST says keeping track of where training data came from helps transparency and accountability. Training data may be protected by copyright.
NIST also says that how generative AI fits with copyright is still being debated in legal settings.
That includes AI-made work that is similar to, but not a copy of, protected work. Using a person's likeness or voice without permission is debated too.
| Statement | True or false? |
|---|---|
| Generative AI can make plagiarism easier. | ? |
| Everyone already agrees exactly how copyright applies to AI-made work. | ? |
| Tracking where training data came from helps accountability. | ? |
| If SH-1 wrote it, nobody else's work could be in it. | ? |
Careful reading, reviewer. Tomorrow is Impact Friday: what a fake does to trust, and how the crew rates two risks.